learning2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "plt.rcParams[\"figure.figsize\"] =(12,9)\n",
    "import os\n",
    "import copy\n",
    "from IPython.display import clear_output\n",
    "import torch\n",
    "from torchvision import transforms\n",
    "import torch.nn as nn\n",
    "import torch.utils.model_zoo as model_zoo\n",
    "from torch.utils.data import DataLoader,Subset\n",
    "from torch.autograd import Variable\n",
    "import PIL\n",
    "import torch.optim as optim\n",
    "import compression as comp\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "scrolled": true
   },
   "source": [
    "# define the dataloader and load the data "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "data is loaded\n",
      "device is cuda:0\n"
     ]
    }
   ],
   "source": [
    "# initialize the dataloader and normalize the data \n",
    "class Dataset:\n",
    "    \"\"\"load the image / encoded object position representation pairs for training and testing data\"\"\"\n",
    "    def __init__(self, path, mode = 'train'):\n",
    "        self.path=path\n",
    "        self.mode=mode\n",
    "    def __getitem__(self, index):\n",
    "        mode = self.mode\n",
    "        if mode=='train':\n",
    "            fname = '/train-%05d.jpg'\n",
    "        elif mode=='test':\n",
    "            fname = '/test-%05d.jpg'\n",
    "            \n",
    "        if mode=='train':\n",
    "            fname1 = 'train-comp-%05d.npy'\n",
    "        elif mode=='test':\n",
    "            fname1 = 'test-comp-%05d.npy'\n",
    "        img = PIL.Image.open(self.path+fname%index)\n",
    "        vect = np.load(self.path+fname1%index)\n",
    "        transform = transforms.Compose([#transforms.Scale((227,227)),\n",
    "                                        transforms.ToTensor(),\n",
    "                                        #transforms.Normalize(mean=[0.5],std=[0.25])\n",
    "                                       ])\n",
    "        img = transform(img)\n",
    "        if mode=='train':\n",
    "            img.requires_grad=True\n",
    "        vect = torch.FloatTensor(np.concatenate(vect)) \n",
    "        return img, vect \n",
    "\n",
    "    def __len__(self):\n",
    "        return len([f for f in os.listdir(self.path) if f.endswith('.jpg')])\n",
    "\n",
    "# Initialize dataset iterators and find gpu if available \n",
    "train_data = Dataset('./data/training/',mode='train')\n",
    "test_data = Dataset('./data/testing/',mode='test')\n",
    "print('data is loaded')\n",
    "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
    "print('device is '+ str(device))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# define the cnn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Net(nn.Module):\n",
    "    \"\"\" this is a hacked up alexnet where the depth of layers is modified to accomodate grayscale wihtout overfitting \"\"\"\n",
    "    def __init__(self):\n",
    "        super(Net, self).__init__()\n",
    "        self.convnet = nn.Sequential(nn.Conv2d(1,32,11,4,2),\n",
    "                                     nn.ReLU(inplace=True),\n",
    "                                     nn.MaxPool2d(3,2,0,1),\n",
    "                                     nn.Conv2d(32,96,5,1,2),\n",
    "                                     nn.ReLU(inplace=True),\n",
    "                                     nn.MaxPool2d(3,2,0,1),\n",
    "                                     nn.Conv2d(96,96*3,3,1,1),\n",
    "                                     nn.ReLU(inplace=True),\n",
    "                                     nn.Conv2d(96*3,96*2,3,1,1),\n",
    "                                     nn.ReLU(inplace=True),\n",
    "                                     nn.Conv2d(96*2,96*2,3,1,1),\n",
    "                                     nn.ReLU(inplace=True),\n",
    "                                     nn.MaxPool2d(3,2,0,1)\n",
    "                                    )\n",
    "        self.fcnet = nn.Sequential(nn.Dropout(),\n",
    "                          nn.Linear(192*3**2,192*3**2//2),\n",
    "                          nn.ReLU(inplace=True),\n",
    "                          nn.Dropout(),\n",
    "                          nn.Linear(192*3**2//2,192*3**2//2),\n",
    "                          nn.ReLU(inplace=True),\n",
    "                          nn.Linear(192*3**2//2,500)\n",
    "                          )\n",
    "    def forward(self, img):\n",
    "        output = self.convnet(img)\n",
    "        output = output.view(-1, 192*3**2)\n",
    "        output = self.fcnet(output)\n",
    "        return output"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# overfit on one sample\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 282,
   "metadata": {},
   "outputs": [],
   "source": [
    "x,y = next(iter(train_loader))\n",
    "x = x.to(device)\n",
    "y = y.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 261,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Net().to(device).train()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 284,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "108.140673828125\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-284-f8242304658d>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     11\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     12\u001b[0m     \u001b[0mloss\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 13\u001b[1;33m     \u001b[0moptimizer\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mstep\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m~\\Miniconda3\\envs\\kevintorch-env\\lib\\site-packages\\torch\\optim\\adam.py\u001b[0m in \u001b[0;36mstep\u001b[1;34m(self, closure)\u001b[0m\n\u001b[0;32m     91\u001b[0m                 \u001b[1;31m# Decay the first and second moment running average coefficient\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     92\u001b[0m                 \u001b[0mexp_avg\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmul_\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mbeta1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0madd_\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m1\u001b[0m \u001b[1;33m-\u001b[0m \u001b[0mbeta1\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mgrad\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 93\u001b[1;33m                 \u001b[0mexp_avg_sq\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmul_\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mbeta2\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0maddcmul_\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m1\u001b[0m \u001b[1;33m-\u001b[0m \u001b[0mbeta2\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mgrad\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mgrad\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     94\u001b[0m                 \u001b[1;32mif\u001b[0m \u001b[0mamsgrad\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     95\u001b[0m                     \u001b[1;31m# Maintains the maximum of all 2nd moment running avg. till now\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "criterion = nn.MSELoss()\n",
    "optimizer = optim.Adam(model.parameters(), lr=0.001)\n",
    "\n",
    "for i in range(50000):\n",
    "    optimizer.zero_grad() # zero gradients\n",
    "    y_ = model(x)\n",
    "    loss = criterion(y_,y)\n",
    "    if i%10==0:\n",
    "        clear_output(wait=True)\n",
    "        print(loss.cpu().data.item()/k)\n",
    "\n",
    "    loss.backward()\n",
    "    optimizer.step()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 276,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = model.eval()\n",
    "y_ = model(x).detach().cpu().numpy()\n",
    "y = y.cpu().numpy()\n",
    "model = model.train()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 280,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x23b67a8dda0>"
      ]
     },
     "execution_count": 280,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x648 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "n = 5 # number to investigate \n",
    "S = np.load('./data/sensing_matrix.npy')\n",
    "shape = (150,150)\n",
    "L = 5\n",
    "M = int(round(np.sqrt(2)*150))\n",
    "pts = comp.decompress(y_[n].reshape(5,100),S,shape,L,M)\n",
    "im = x[n].cpu().detach().numpy().squeeze()\n",
    "for p in pts:\n",
    "    im=comp.draw_plus(p,im)\n",
    "plt.imshow(im)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x23ab1ad2748>]"
      ]
     },
     "execution_count": 226,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x648 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "n = 33\n",
    "plt.plot(y[n])\n",
    "plt.plot(y_[n])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "metadata": {},
   "outputs": [],
   "source": [
    "# totally works on 1 sample. it will work on all of them. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# train the network "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 285,
   "metadata": {},
   "outputs": [],
   "source": [
    "# initialize the model \n",
    "model = Net().to(device).train()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 288,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 0 completed in 27.11 seconds with avg training loss 324.93\n",
      "model saved (early stop).  validation loss was  816.38599609375\n",
      "epoch 1 completed in 26.35 seconds with avg training loss 322.4\n",
      "epoch 2 completed in 26.47 seconds with avg training loss 322.97\n",
      "epoch 3 completed in 26.66 seconds with avg training loss 319.76\n",
      "epoch 4 completed in 26.06 seconds with avg training loss 319.61\n",
      "epoch 5 completed in 26.34 seconds with avg training loss 320.06\n",
      "model not saved. validation loss was  877.97365234375\n",
      "epoch 6 completed in 25.78 seconds with avg training loss 319.82\n",
      "epoch 7 completed in 26.19 seconds with avg training loss 318.73\n",
      "epoch 8 completed in 25.77 seconds with avg training loss 318.8\n",
      "epoch 9 completed in 26.28 seconds with avg training loss 316.93\n",
      "epoch 10 completed in 25.95 seconds with avg training loss 319.02\n",
      "model saved (early stop).  validation loss was  812.7486132812501\n",
      "epoch 11 completed in 26.17 seconds with avg training loss 316.71\n",
      "epoch 12 completed in 26.08 seconds with avg training loss 317.28\n",
      "epoch 13 completed in 26.38 seconds with avg training loss 315.72\n",
      "epoch 14 completed in 25.83 seconds with avg training loss 314.7\n",
      "epoch 15 completed in 26.17 seconds with avg training loss 315.49\n",
      "model not saved. validation loss was  815.15966796875\n",
      "epoch 16 completed in 25.94 seconds with avg training loss 316.31\n",
      "epoch 17 completed in 26.31 seconds with avg training loss 314.42\n",
      "epoch 18 completed in 26.13 seconds with avg training loss 315.68\n",
      "epoch 19 completed in 26.05 seconds with avg training loss 313.65\n",
      "epoch 20 completed in 25.73 seconds with avg training loss 312.73\n",
      "model not saved. validation loss was  928.2956054687501\n",
      "epoch 21 completed in 25.64 seconds with avg training loss 314.24\n",
      "epoch 22 completed in 26.38 seconds with avg training loss 312.81\n",
      "epoch 23 completed in 25.98 seconds with avg training loss 313.39\n",
      "epoch 24 completed in 26.55 seconds with avg training loss 312.1\n",
      "epoch 25 completed in 26.33 seconds with avg training loss 310.81\n",
      "model not saved. validation loss was  825.205546875\n",
      "epoch 26 completed in 26.48 seconds with avg training loss 311.87\n",
      "epoch 27 completed in 26.91 seconds with avg training loss 310.93\n",
      "epoch 28 completed in 26.92 seconds with avg training loss 310.19\n",
      "epoch 29 completed in 26.5 seconds with avg training loss 310.56\n",
      "epoch 30 completed in 26.91 seconds with avg training loss 310.53\n",
      "model not saved. validation loss was  814.1932226562501\n",
      "epoch 31 completed in 26.86 seconds with avg training loss 309.54\n",
      "epoch 32 completed in 26.5 seconds with avg training loss 308.85\n",
      "epoch 33 completed in 27.09 seconds with avg training loss 310.07\n",
      "epoch 34 completed in 26.59 seconds with avg training loss 308.78\n",
      "epoch 35 completed in 27.05 seconds with avg training loss 309.44\n",
      "model saved (early stop).  validation loss was  768.2977734374999\n",
      "epoch 36 completed in 27.19 seconds with avg training loss 309.07\n",
      "epoch 37 completed in 26.94 seconds with avg training loss 308.94\n",
      "epoch 38 completed in 26.67 seconds with avg training loss 308.76\n",
      "epoch 39 completed in 26.5 seconds with avg training loss 308.56\n",
      "epoch 40 completed in 27.08 seconds with avg training loss 308.79\n",
      "model not saved. validation loss was  826.905546875\n",
      "epoch 41 completed in 26.52 seconds with avg training loss 308.54\n",
      "epoch 42 completed in 27.0 seconds with avg training loss 308.33\n",
      "epoch 43 completed in 27.13 seconds with avg training loss 308.74\n",
      "epoch 44 completed in 27.14 seconds with avg training loss 310.09\n",
      "epoch 45 completed in 27.0 seconds with avg training loss 308.86\n",
      "model not saved. validation loss was  808.4614453125\n",
      "epoch 46 completed in 27.47 seconds with avg training loss 308.46\n",
      "epoch 47 completed in 26.58 seconds with avg training loss 308.61\n",
      "epoch 48 completed in 26.81 seconds with avg training loss 307.83\n",
      "epoch 49 completed in 27.22 seconds with avg training loss 308.52\n",
      "epoch 50 completed in 27.09 seconds with avg training loss 308.19\n",
      "model not saved. validation loss was  790.310625\n",
      "epoch 51 completed in 27.2 seconds with avg training loss 308.47\n",
      "epoch 52 completed in 26.88 seconds with avg training loss 308.62\n",
      "epoch 53 completed in 26.77 seconds with avg training loss 308.66\n",
      "epoch 54 completed in 26.48 seconds with avg training loss 308.07\n",
      "epoch 55 completed in 26.27 seconds with avg training loss 308.03\n",
      "model not saved. validation loss was  815.295703125\n",
      "epoch 56 completed in 26.3 seconds with avg training loss 308.52\n",
      "epoch 57 completed in 27.36 seconds with avg training loss 307.73\n",
      "epoch 58 completed in 26.94 seconds with avg training loss 309.28\n",
      "epoch 59 completed in 27.3 seconds with avg training loss 308.09\n",
      "epoch 60 completed in 27.19 seconds with avg training loss 308.09\n",
      "model not saved. validation loss was  805.0216601562499\n",
      "epoch 61 completed in 27.0 seconds with avg training loss 307.92\n",
      "epoch 62 completed in 27.03 seconds with avg training loss 307.91\n",
      "epoch 63 completed in 27.16 seconds with avg training loss 307.86\n",
      "epoch 64 completed in 26.78 seconds with avg training loss 307.87\n",
      "epoch 65 completed in 26.83 seconds with avg training loss 307.45\n",
      "model not saved. validation loss was  819.4749999999999\n",
      "epoch 66 completed in 26.42 seconds with avg training loss 308.27\n",
      "epoch 67 completed in 26.55 seconds with avg training loss 309.25\n",
      "epoch 68 completed in 26.58 seconds with avg training loss 307.98\n",
      "epoch 69 completed in 26.8 seconds with avg training loss 308.72\n",
      "epoch 70 completed in 26.48 seconds with avg training loss 307.38\n",
      "model not saved. validation loss was  831.0655468750001\n",
      "epoch 71 completed in 26.8 seconds with avg training loss 308.66\n",
      "epoch 72 completed in 26.8 seconds with avg training loss 307.45\n",
      "epoch 73 completed in 26.97 seconds with avg training loss 307.41\n",
      "epoch 74 completed in 26.86 seconds with avg training loss 307.93\n",
      "epoch 75 completed in 26.61 seconds with avg training loss 307.5\n",
      "model not saved. validation loss was  820.4373828124999\n",
      "epoch 76 completed in 26.94 seconds with avg training loss 307.99\n",
      "epoch 77 completed in 26.59 seconds with avg training loss 307.52\n",
      "epoch 78 completed in 26.89 seconds with avg training loss 307.91\n",
      "epoch 79 completed in 26.73 seconds with avg training loss 309.14\n",
      "epoch 80 completed in 26.81 seconds with avg training loss 307.86\n",
      "model not saved. validation loss was  784.691015625\n",
      "epoch 81 completed in 26.89 seconds with avg training loss 307.83\n",
      "Epoch    81: reducing learning rate of group 0 to 1.0000e-04.\n",
      "epoch 82 completed in 26.61 seconds with avg training loss 307.11\n",
      "epoch 83 completed in 27.16 seconds with avg training loss 306.93\n",
      "epoch 84 completed in 27.36 seconds with avg training loss 307.41\n",
      "epoch 85 completed in 26.97 seconds with avg training loss 306.48\n",
      "model not saved. validation loss was  804.177734375\n",
      "epoch 86 completed in 26.66 seconds with avg training loss 306.68\n",
      "epoch 87 completed in 27.09 seconds with avg training loss 306.54\n",
      "epoch 88 completed in 26.78 seconds with avg training loss 306.61\n",
      "epoch 89 completed in 26.44 seconds with avg training loss 306.7\n",
      "epoch 90 completed in 27.03 seconds with avg training loss 306.52\n",
      "model not saved. validation loss was  799.4468359375\n",
      "epoch 91 completed in 26.92 seconds with avg training loss 306.82\n",
      "epoch 92 completed in 26.72 seconds with avg training loss 306.13\n",
      "epoch 93 completed in 26.52 seconds with avg training loss 306.57\n",
      "epoch 94 completed in 26.64 seconds with avg training loss 306.52\n",
      "epoch 95 completed in 26.83 seconds with avg training loss 305.96\n",
      "model not saved. validation loss was  808.2717578125\n",
      "epoch 96 completed in 26.64 seconds with avg training loss 306.58\n",
      "epoch 97 completed in 27.02 seconds with avg training loss 305.86\n",
      "epoch 98 completed in 26.8 seconds with avg training loss 305.56\n",
      "epoch 99 completed in 150.14 seconds with avg training loss 306.34\n",
      "epoch 100 completed in 52.72 seconds with avg training loss 306.59\n",
      "model not saved. validation loss was  822.2645898437501\n",
      "epoch 101 completed in 26.8 seconds with avg training loss 306.56\n",
      "epoch 102 completed in 26.5 seconds with avg training loss 306.2\n",
      "epoch 103 completed in 26.78 seconds with avg training loss 306.16\n",
      "epoch 104 completed in 26.48 seconds with avg training loss 305.95\n",
      "epoch 105 completed in 26.52 seconds with avg training loss 307.1\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model not saved. validation loss was  788.655\n",
      "epoch 106 completed in 26.78 seconds with avg training loss 306.07\n",
      "epoch 107 completed in 26.61 seconds with avg training loss 306.2\n",
      "epoch 108 completed in 27.02 seconds with avg training loss 306.44\n",
      "epoch 109 completed in 26.45 seconds with avg training loss 305.9\n",
      "Epoch   109: reducing learning rate of group 0 to 1.0000e-05.\n",
      "epoch 110 completed in 26.56 seconds with avg training loss 305.93\n",
      "model not saved. validation loss was  800.3966601562499\n",
      "epoch 111 completed in 26.44 seconds with avg training loss 306.61\n",
      "epoch 112 completed in 26.69 seconds with avg training loss 305.75\n",
      "epoch 113 completed in 26.63 seconds with avg training loss 306.1\n",
      "epoch 114 completed in 26.73 seconds with avg training loss 306.2\n",
      "epoch 115 completed in 26.58 seconds with avg training loss 306.41\n",
      "model not saved. validation loss was  806.1481640625\n",
      "epoch 116 completed in 26.77 seconds with avg training loss 306.85\n",
      "epoch 117 completed in 26.58 seconds with avg training loss 306.59\n",
      "epoch 118 completed in 26.41 seconds with avg training loss 306.33\n",
      "epoch 119 completed in 26.92 seconds with avg training loss 306.27\n",
      "epoch 120 completed in 26.69 seconds with avg training loss 306.37\n",
      "Epoch   120: reducing learning rate of group 0 to 1.0000e-06.\n",
      "model not saved. validation loss was  800.1089843750001\n",
      "epoch 121 completed in 26.84 seconds with avg training loss 306.27\n",
      "epoch 122 completed in 26.92 seconds with avg training loss 306.68\n",
      "epoch 123 completed in 26.92 seconds with avg training loss 306.17\n",
      "epoch 124 completed in 26.48 seconds with avg training loss 306.29\n",
      "epoch 125 completed in 26.83 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  809.9961523437499\n",
      "epoch 126 completed in 26.42 seconds with avg training loss 306.53\n",
      "epoch 127 completed in 26.69 seconds with avg training loss 306.5\n",
      "epoch 128 completed in 26.53 seconds with avg training loss 305.79\n",
      "epoch 129 completed in 26.77 seconds with avg training loss 306.22\n",
      "epoch 130 completed in 26.95 seconds with avg training loss 305.79\n",
      "model not saved. validation loss was  805.7855468749999\n",
      "epoch 131 completed in 26.78 seconds with avg training loss 305.72\n",
      "Epoch   131: reducing learning rate of group 0 to 1.0000e-07.\n",
      "epoch 132 completed in 26.75 seconds with avg training loss 306.45\n",
      "epoch 133 completed in 26.42 seconds with avg training loss 306.63\n",
      "epoch 134 completed in 26.66 seconds with avg training loss 306.22\n",
      "epoch 135 completed in 26.77 seconds with avg training loss 306.3\n",
      "model not saved. validation loss was  810.75865234375\n",
      "epoch 136 completed in 26.81 seconds with avg training loss 305.75\n",
      "epoch 137 completed in 26.61 seconds with avg training loss 306.07\n",
      "epoch 138 completed in 26.69 seconds with avg training loss 306.52\n",
      "epoch 139 completed in 26.53 seconds with avg training loss 306.39\n",
      "epoch 140 completed in 26.44 seconds with avg training loss 306.65\n",
      "model not saved. validation loss was  812.2453125000001\n",
      "epoch 141 completed in 27.0 seconds with avg training loss 306.61\n",
      "epoch 142 completed in 26.69 seconds with avg training loss 306.2\n",
      "Epoch   142: reducing learning rate of group 0 to 1.0000e-08.\n",
      "epoch 143 completed in 26.59 seconds with avg training loss 306.15\n",
      "epoch 144 completed in 26.84 seconds with avg training loss 306.01\n",
      "epoch 145 completed in 26.81 seconds with avg training loss 306.25\n",
      "model not saved. validation loss was  804.45052734375\n",
      "epoch 146 completed in 26.75 seconds with avg training loss 306.21\n",
      "epoch 147 completed in 27.39 seconds with avg training loss 305.85\n",
      "epoch 148 completed in 26.89 seconds with avg training loss 306.07\n",
      "epoch 149 completed in 26.48 seconds with avg training loss 306.2\n",
      "epoch 150 completed in 26.72 seconds with avg training loss 305.72\n",
      "model not saved. validation loss was  801.3450390625001\n",
      "epoch 151 completed in 26.41 seconds with avg training loss 306.03\n",
      "epoch 152 completed in 26.77 seconds with avg training loss 305.73\n",
      "epoch 153 completed in 26.73 seconds with avg training loss 306.32\n",
      "epoch 154 completed in 26.84 seconds with avg training loss 305.73\n",
      "epoch 155 completed in 27.19 seconds with avg training loss 306.42\n",
      "model not saved. validation loss was  806.694296875\n",
      "epoch 156 completed in 26.95 seconds with avg training loss 305.83\n",
      "epoch 157 completed in 26.91 seconds with avg training loss 306.8\n",
      "epoch 158 completed in 26.94 seconds with avg training loss 305.98\n",
      "epoch 159 completed in 26.89 seconds with avg training loss 306.16\n",
      "epoch 160 completed in 27.0 seconds with avg training loss 306.44\n",
      "model not saved. validation loss was  806.3647265625\n",
      "epoch 161 completed in 27.11 seconds with avg training loss 306.11\n",
      "epoch 162 completed in 27.08 seconds with avg training loss 305.98\n",
      "epoch 163 completed in 26.92 seconds with avg training loss 306.87\n",
      "epoch 164 completed in 27.13 seconds with avg training loss 306.37\n",
      "epoch 165 completed in 26.48 seconds with avg training loss 305.95\n",
      "model not saved. validation loss was  803.5670507812501\n",
      "epoch 166 completed in 26.38 seconds with avg training loss 306.55\n",
      "epoch 167 completed in 26.64 seconds with avg training loss 305.84\n",
      "epoch 168 completed in 26.52 seconds with avg training loss 305.98\n",
      "epoch 169 completed in 26.91 seconds with avg training loss 306.2\n",
      "epoch 170 completed in 26.34 seconds with avg training loss 306.13\n",
      "model not saved. validation loss was  796.11458984375\n",
      "epoch 171 completed in 27.14 seconds with avg training loss 306.61\n",
      "epoch 172 completed in 26.66 seconds with avg training loss 306.2\n",
      "epoch 173 completed in 26.52 seconds with avg training loss 306.24\n",
      "epoch 174 completed in 26.77 seconds with avg training loss 305.86\n",
      "epoch 175 completed in 26.83 seconds with avg training loss 306.12\n",
      "model not saved. validation loss was  811.5255468749999\n",
      "epoch 176 completed in 26.44 seconds with avg training loss 306.51\n",
      "epoch 177 completed in 27.27 seconds with avg training loss 306.03\n",
      "epoch 178 completed in 27.28 seconds with avg training loss 306.43\n",
      "epoch 179 completed in 26.98 seconds with avg training loss 306.35\n",
      "epoch 180 completed in 26.98 seconds with avg training loss 305.89\n",
      "model not saved. validation loss was  804.78568359375\n",
      "epoch 181 completed in 27.28 seconds with avg training loss 306.34\n",
      "epoch 182 completed in 27.2 seconds with avg training loss 306.42\n",
      "epoch 183 completed in 27.27 seconds with avg training loss 306.3\n",
      "epoch 184 completed in 26.39 seconds with avg training loss 306.43\n",
      "epoch 185 completed in 26.36 seconds with avg training loss 306.47\n",
      "model not saved. validation loss was  804.5201367187499\n",
      "epoch 186 completed in 26.59 seconds with avg training loss 305.97\n",
      "epoch 187 completed in 26.48 seconds with avg training loss 306.35\n",
      "epoch 188 completed in 26.92 seconds with avg training loss 306.62\n",
      "epoch 189 completed in 26.81 seconds with avg training loss 306.18\n",
      "epoch 190 completed in 27.03 seconds with avg training loss 306.28\n",
      "model not saved. validation loss was  806.1422070312499\n",
      "epoch 191 completed in 27.3 seconds with avg training loss 306.31\n",
      "epoch 192 completed in 26.97 seconds with avg training loss 306.84\n",
      "epoch 193 completed in 26.75 seconds with avg training loss 305.99\n",
      "epoch 194 completed in 27.41 seconds with avg training loss 305.39\n",
      "epoch 195 completed in 26.94 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  805.190078125\n",
      "epoch 196 completed in 26.8 seconds with avg training loss 306.49\n",
      "epoch 197 completed in 26.63 seconds with avg training loss 306.56\n",
      "epoch 198 completed in 26.88 seconds with avg training loss 306.36\n",
      "epoch 199 completed in 27.0 seconds with avg training loss 306.05\n",
      "epoch 200 completed in 27.39 seconds with avg training loss 306.29\n",
      "model not saved. validation loss was  809.2582226562499\n",
      "epoch 201 completed in 27.52 seconds with avg training loss 306.11\n",
      "epoch 202 completed in 27.28 seconds with avg training loss 306.22\n",
      "epoch 203 completed in 27.25 seconds with avg training loss 306.27\n",
      "epoch 204 completed in 27.03 seconds with avg training loss 306.01\n",
      "epoch 205 completed in 27.13 seconds with avg training loss 306.36\n",
      "model not saved. validation loss was  809.1823828125\n",
      "epoch 206 completed in 26.98 seconds with avg training loss 305.92\n",
      "epoch 207 completed in 27.19 seconds with avg training loss 305.78\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 208 completed in 26.77 seconds with avg training loss 306.13\n",
      "epoch 209 completed in 26.92 seconds with avg training loss 306.06\n",
      "epoch 210 completed in 26.72 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  810.40435546875\n",
      "epoch 211 completed in 26.84 seconds with avg training loss 306.0\n",
      "epoch 212 completed in 26.66 seconds with avg training loss 305.88\n",
      "epoch 213 completed in 26.98 seconds with avg training loss 306.45\n",
      "epoch 214 completed in 27.3 seconds with avg training loss 306.52\n",
      "epoch 215 completed in 26.83 seconds with avg training loss 306.32\n",
      "model not saved. validation loss was  799.7993945312501\n",
      "epoch 216 completed in 26.77 seconds with avg training loss 306.6\n",
      "epoch 217 completed in 26.89 seconds with avg training loss 306.28\n",
      "epoch 218 completed in 26.5 seconds with avg training loss 306.14\n",
      "epoch 219 completed in 26.36 seconds with avg training loss 305.82\n",
      "epoch 220 completed in 26.91 seconds with avg training loss 305.82\n",
      "model not saved. validation loss was  804.48892578125\n",
      "epoch 221 completed in 26.97 seconds with avg training loss 306.08\n",
      "epoch 222 completed in 27.09 seconds with avg training loss 306.34\n",
      "epoch 223 completed in 27.0 seconds with avg training loss 306.24\n",
      "epoch 224 completed in 26.94 seconds with avg training loss 306.33\n",
      "epoch 225 completed in 27.0 seconds with avg training loss 306.84\n",
      "model not saved. validation loss was  802.15232421875\n",
      "epoch 226 completed in 26.94 seconds with avg training loss 305.6\n",
      "epoch 227 completed in 26.91 seconds with avg training loss 306.21\n",
      "epoch 228 completed in 27.05 seconds with avg training loss 305.81\n",
      "epoch 229 completed in 26.88 seconds with avg training loss 305.8\n",
      "epoch 230 completed in 27.22 seconds with avg training loss 306.5\n",
      "model not saved. validation loss was  805.2492578125\n",
      "epoch 231 completed in 26.61 seconds with avg training loss 305.81\n",
      "epoch 232 completed in 27.42 seconds with avg training loss 306.0\n",
      "epoch 233 completed in 27.17 seconds with avg training loss 306.34\n",
      "epoch 234 completed in 27.05 seconds with avg training loss 305.93\n",
      "epoch 235 completed in 26.94 seconds with avg training loss 305.91\n",
      "model not saved. validation loss was  808.491171875\n",
      "epoch 236 completed in 26.44 seconds with avg training loss 306.38\n",
      "epoch 237 completed in 27.17 seconds with avg training loss 306.14\n",
      "epoch 238 completed in 26.81 seconds with avg training loss 306.27\n",
      "epoch 239 completed in 26.95 seconds with avg training loss 306.13\n",
      "epoch 240 completed in 27.16 seconds with avg training loss 306.45\n",
      "model not saved. validation loss was  808.9234765625\n",
      "epoch 241 completed in 26.36 seconds with avg training loss 306.13\n",
      "epoch 242 completed in 27.36 seconds with avg training loss 306.19\n",
      "epoch 243 completed in 26.77 seconds with avg training loss 306.52\n",
      "epoch 244 completed in 26.84 seconds with avg training loss 305.97\n",
      "epoch 245 completed in 27.06 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  806.2317578125\n",
      "epoch 246 completed in 26.84 seconds with avg training loss 305.83\n",
      "epoch 247 completed in 27.05 seconds with avg training loss 306.15\n",
      "epoch 248 completed in 26.58 seconds with avg training loss 306.1\n",
      "epoch 249 completed in 26.7 seconds with avg training loss 306.06\n",
      "epoch 250 completed in 26.38 seconds with avg training loss 306.27\n",
      "model not saved. validation loss was  801.516640625\n",
      "epoch 251 completed in 26.8 seconds with avg training loss 306.5\n",
      "epoch 252 completed in 26.89 seconds with avg training loss 306.12\n",
      "epoch 253 completed in 26.5 seconds with avg training loss 305.95\n",
      "epoch 254 completed in 26.83 seconds with avg training loss 306.22\n",
      "epoch 255 completed in 26.91 seconds with avg training loss 306.04\n",
      "model not saved. validation loss was  812.2918359375001\n",
      "epoch 256 completed in 26.95 seconds with avg training loss 306.0\n",
      "epoch 257 completed in 26.86 seconds with avg training loss 305.95\n",
      "epoch 258 completed in 26.83 seconds with avg training loss 306.55\n",
      "epoch 259 completed in 27.61 seconds with avg training loss 305.91\n",
      "epoch 260 completed in 26.38 seconds with avg training loss 306.15\n",
      "model not saved. validation loss was  809.66556640625\n",
      "epoch 261 completed in 26.92 seconds with avg training loss 306.64\n",
      "epoch 262 completed in 26.8 seconds with avg training loss 305.65\n",
      "epoch 263 completed in 26.97 seconds with avg training loss 306.21\n",
      "epoch 264 completed in 26.81 seconds with avg training loss 306.63\n",
      "epoch 265 completed in 26.77 seconds with avg training loss 306.38\n",
      "model not saved. validation loss was  797.7723828124999\n",
      "epoch 266 completed in 26.67 seconds with avg training loss 306.55\n",
      "epoch 267 completed in 26.19 seconds with avg training loss 306.56\n",
      "epoch 268 completed in 26.92 seconds with avg training loss 305.9\n",
      "epoch 269 completed in 26.81 seconds with avg training loss 305.78\n",
      "epoch 270 completed in 26.47 seconds with avg training loss 305.92\n",
      "model not saved. validation loss was  810.2747265625\n",
      "epoch 271 completed in 27.16 seconds with avg training loss 305.89\n",
      "epoch 272 completed in 26.64 seconds with avg training loss 305.8\n",
      "epoch 273 completed in 26.58 seconds with avg training loss 306.18\n",
      "epoch 274 completed in 26.77 seconds with avg training loss 306.22\n",
      "epoch 275 completed in 27.0 seconds with avg training loss 306.23\n",
      "model not saved. validation loss was  805.583125\n",
      "epoch 276 completed in 26.97 seconds with avg training loss 306.5\n",
      "epoch 277 completed in 26.59 seconds with avg training loss 306.33\n",
      "epoch 278 completed in 26.72 seconds with avg training loss 306.03\n",
      "epoch 279 completed in 27.38 seconds with avg training loss 305.98\n",
      "epoch 280 completed in 26.2 seconds with avg training loss 306.47\n",
      "model not saved. validation loss was  808.0695312500001\n",
      "epoch 281 completed in 26.72 seconds with avg training loss 306.29\n",
      "epoch 282 completed in 26.67 seconds with avg training loss 306.1\n",
      "epoch 283 completed in 26.73 seconds with avg training loss 305.75\n",
      "epoch 284 completed in 26.77 seconds with avg training loss 306.11\n",
      "epoch 285 completed in 26.88 seconds with avg training loss 306.13\n",
      "model not saved. validation loss was  809.5183984375\n",
      "epoch 286 completed in 26.95 seconds with avg training loss 305.99\n",
      "epoch 287 completed in 26.53 seconds with avg training loss 306.55\n",
      "epoch 288 completed in 26.78 seconds with avg training loss 305.69\n",
      "epoch 289 completed in 26.52 seconds with avg training loss 306.15\n",
      "epoch 290 completed in 26.69 seconds with avg training loss 306.49\n",
      "model not saved. validation loss was  804.8846093750001\n",
      "epoch 291 completed in 27.02 seconds with avg training loss 306.21\n",
      "epoch 292 completed in 26.94 seconds with avg training loss 306.52\n",
      "epoch 293 completed in 26.64 seconds with avg training loss 306.5\n",
      "epoch 294 completed in 27.19 seconds with avg training loss 305.89\n",
      "epoch 295 completed in 27.17 seconds with avg training loss 306.07\n",
      "model not saved. validation loss was  802.03150390625\n",
      "epoch 296 completed in 27.25 seconds with avg training loss 306.67\n",
      "epoch 297 completed in 27.03 seconds with avg training loss 306.21\n",
      "epoch 298 completed in 26.69 seconds with avg training loss 306.06\n",
      "epoch 299 completed in 26.97 seconds with avg training loss 305.91\n",
      "epoch 300 completed in 26.88 seconds with avg training loss 307.0\n",
      "model not saved. validation loss was  804.6343554687501\n",
      "epoch 301 completed in 27.14 seconds with avg training loss 306.17\n",
      "epoch 302 completed in 26.77 seconds with avg training loss 306.24\n",
      "epoch 303 completed in 26.88 seconds with avg training loss 306.17\n",
      "epoch 304 completed in 27.28 seconds with avg training loss 306.31\n",
      "epoch 305 completed in 26.94 seconds with avg training loss 305.78\n",
      "model not saved. validation loss was  806.57498046875\n",
      "epoch 306 completed in 27.31 seconds with avg training loss 305.99\n",
      "epoch 307 completed in 27.2 seconds with avg training loss 306.71\n",
      "epoch 308 completed in 27.09 seconds with avg training loss 305.92\n",
      "epoch 309 completed in 26.98 seconds with avg training loss 306.16\n",
      "epoch 310 completed in 27.19 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  806.21806640625\n",
      "epoch 311 completed in 26.83 seconds with avg training loss 305.89\n",
      "epoch 312 completed in 27.13 seconds with avg training loss 305.92\n",
      "epoch 313 completed in 27.22 seconds with avg training loss 306.47\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 314 completed in 27.05 seconds with avg training loss 306.14\n",
      "epoch 315 completed in 27.28 seconds with avg training loss 306.5\n",
      "model not saved. validation loss was  805.9923632812499\n",
      "epoch 316 completed in 27.22 seconds with avg training loss 305.99\n",
      "epoch 317 completed in 27.03 seconds with avg training loss 305.81\n",
      "epoch 318 completed in 26.91 seconds with avg training loss 305.8\n",
      "epoch 319 completed in 26.97 seconds with avg training loss 306.37\n",
      "epoch 320 completed in 26.98 seconds with avg training loss 306.04\n",
      "model not saved. validation loss was  800.0200390625\n",
      "epoch 321 completed in 26.88 seconds with avg training loss 306.6\n",
      "epoch 322 completed in 26.67 seconds with avg training loss 305.68\n",
      "epoch 323 completed in 27.0 seconds with avg training loss 306.27\n",
      "epoch 324 completed in 26.84 seconds with avg training loss 306.37\n",
      "epoch 325 completed in 27.08 seconds with avg training loss 305.99\n",
      "model not saved. validation loss was  801.35544921875\n",
      "epoch 326 completed in 27.75 seconds with avg training loss 306.72\n",
      "epoch 327 completed in 27.06 seconds with avg training loss 305.79\n",
      "epoch 328 completed in 27.45 seconds with avg training loss 306.09\n",
      "epoch 329 completed in 27.06 seconds with avg training loss 305.99\n",
      "epoch 330 completed in 27.0 seconds with avg training loss 306.46\n",
      "model not saved. validation loss was  804.1344726562501\n",
      "epoch 331 completed in 26.92 seconds with avg training loss 306.37\n",
      "epoch 332 completed in 26.91 seconds with avg training loss 306.43\n",
      "epoch 333 completed in 26.92 seconds with avg training loss 306.42\n",
      "epoch 334 completed in 26.8 seconds with avg training loss 305.8\n",
      "epoch 335 completed in 27.05 seconds with avg training loss 305.99\n",
      "model not saved. validation loss was  807.725\n",
      "epoch 336 completed in 27.31 seconds with avg training loss 305.7\n",
      "epoch 337 completed in 27.05 seconds with avg training loss 306.23\n",
      "epoch 338 completed in 26.92 seconds with avg training loss 305.86\n",
      "epoch 339 completed in 27.48 seconds with avg training loss 305.94\n",
      "epoch 340 completed in 26.72 seconds with avg training loss 305.97\n",
      "model not saved. validation loss was  806.632109375\n",
      "epoch 341 completed in 26.56 seconds with avg training loss 305.97\n",
      "epoch 342 completed in 26.73 seconds with avg training loss 306.05\n",
      "epoch 343 completed in 26.72 seconds with avg training loss 306.49\n",
      "epoch 344 completed in 26.84 seconds with avg training loss 306.17\n",
      "epoch 345 completed in 27.27 seconds with avg training loss 306.18\n",
      "model not saved. validation loss was  806.85876953125\n",
      "epoch 346 completed in 27.28 seconds with avg training loss 305.54\n",
      "epoch 347 completed in 27.25 seconds with avg training loss 306.13\n",
      "epoch 348 completed in 26.75 seconds with avg training loss 306.14\n",
      "epoch 349 completed in 26.95 seconds with avg training loss 306.32\n",
      "epoch 350 completed in 26.42 seconds with avg training loss 306.05\n",
      "model not saved. validation loss was  809.83111328125\n",
      "epoch 351 completed in 26.58 seconds with avg training loss 306.42\n",
      "epoch 352 completed in 27.13 seconds with avg training loss 306.76\n",
      "epoch 353 completed in 26.88 seconds with avg training loss 306.58\n",
      "epoch 354 completed in 26.89 seconds with avg training loss 306.5\n",
      "epoch 355 completed in 27.17 seconds with avg training loss 306.33\n",
      "model not saved. validation loss was  806.1885546875001\n",
      "epoch 356 completed in 26.84 seconds with avg training loss 306.27\n",
      "epoch 357 completed in 26.94 seconds with avg training loss 306.17\n",
      "epoch 358 completed in 26.88 seconds with avg training loss 306.51\n",
      "epoch 359 completed in 26.8 seconds with avg training loss 306.67\n",
      "epoch 360 completed in 26.72 seconds with avg training loss 306.25\n",
      "model not saved. validation loss was  799.3047656250001\n",
      "epoch 361 completed in 26.59 seconds with avg training loss 306.28\n",
      "epoch 362 completed in 26.77 seconds with avg training loss 306.36\n",
      "epoch 363 completed in 26.88 seconds with avg training loss 305.89\n",
      "epoch 364 completed in 27.0 seconds with avg training loss 306.24\n",
      "epoch 365 completed in 26.7 seconds with avg training loss 305.79\n",
      "model not saved. validation loss was  806.1847265624999\n",
      "epoch 366 completed in 26.7 seconds with avg training loss 306.79\n",
      "epoch 367 completed in 26.7 seconds with avg training loss 306.45\n",
      "epoch 368 completed in 26.38 seconds with avg training loss 306.15\n",
      "epoch 369 completed in 26.48 seconds with avg training loss 305.67\n",
      "epoch 370 completed in 26.73 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  804.8671484375\n",
      "epoch 371 completed in 26.52 seconds with avg training loss 306.12\n",
      "epoch 372 completed in 26.69 seconds with avg training loss 306.01\n",
      "epoch 373 completed in 26.89 seconds with avg training loss 306.35\n",
      "epoch 374 completed in 26.81 seconds with avg training loss 306.12\n",
      "epoch 375 completed in 26.63 seconds with avg training loss 305.93\n",
      "model not saved. validation loss was  805.42228515625\n",
      "epoch 376 completed in 27.13 seconds with avg training loss 306.03\n",
      "epoch 377 completed in 27.14 seconds with avg training loss 305.95\n",
      "epoch 378 completed in 26.86 seconds with avg training loss 306.85\n",
      "epoch 379 completed in 26.58 seconds with avg training loss 306.27\n",
      "epoch 380 completed in 26.97 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  811.39365234375\n",
      "epoch 381 completed in 27.22 seconds with avg training loss 305.9\n",
      "epoch 382 completed in 27.14 seconds with avg training loss 306.49\n",
      "epoch 383 completed in 26.73 seconds with avg training loss 306.2\n",
      "epoch 384 completed in 27.27 seconds with avg training loss 306.36\n",
      "epoch 385 completed in 27.23 seconds with avg training loss 306.3\n",
      "model not saved. validation loss was  800.1331250000001\n",
      "epoch 386 completed in 26.59 seconds with avg training loss 306.67\n",
      "epoch 387 completed in 27.22 seconds with avg training loss 306.25\n",
      "epoch 388 completed in 27.44 seconds with avg training loss 305.9\n",
      "epoch 389 completed in 26.89 seconds with avg training loss 306.55\n",
      "epoch 390 completed in 27.22 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  800.70236328125\n",
      "epoch 391 completed in 27.17 seconds with avg training loss 305.65\n",
      "epoch 392 completed in 27.31 seconds with avg training loss 306.32\n",
      "epoch 393 completed in 27.27 seconds with avg training loss 305.93\n",
      "epoch 394 completed in 27.2 seconds with avg training loss 306.38\n",
      "epoch 395 completed in 26.86 seconds with avg training loss 306.55\n",
      "model not saved. validation loss was  802.6038281250001\n",
      "epoch 396 completed in 27.0 seconds with avg training loss 306.54\n",
      "epoch 397 completed in 26.56 seconds with avg training loss 306.19\n",
      "epoch 398 completed in 26.94 seconds with avg training loss 306.37\n",
      "epoch 399 completed in 26.73 seconds with avg training loss 306.17\n",
      "epoch 400 completed in 26.73 seconds with avg training loss 306.36\n",
      "model not saved. validation loss was  808.5968164062499\n",
      "epoch 401 completed in 26.67 seconds with avg training loss 305.81\n",
      "epoch 402 completed in 26.66 seconds with avg training loss 307.0\n",
      "epoch 403 completed in 27.34 seconds with avg training loss 305.84\n",
      "epoch 404 completed in 27.72 seconds with avg training loss 306.09\n",
      "epoch 405 completed in 26.73 seconds with avg training loss 306.46\n",
      "model not saved. validation loss was  799.4243554687499\n",
      "epoch 406 completed in 26.7 seconds with avg training loss 306.61\n",
      "epoch 407 completed in 27.16 seconds with avg training loss 306.5\n",
      "epoch 408 completed in 26.83 seconds with avg training loss 306.21\n",
      "epoch 409 completed in 26.94 seconds with avg training loss 306.34\n",
      "epoch 410 completed in 26.95 seconds with avg training loss 306.18\n",
      "model not saved. validation loss was  802.6190624999999\n",
      "epoch 411 completed in 27.23 seconds with avg training loss 306.71\n",
      "epoch 412 completed in 27.0 seconds with avg training loss 306.34\n",
      "epoch 413 completed in 26.94 seconds with avg training loss 305.68\n",
      "epoch 414 completed in 26.66 seconds with avg training loss 305.67\n",
      "epoch 415 completed in 26.83 seconds with avg training loss 306.59\n",
      "model not saved. validation loss was  809.80435546875\n",
      "epoch 416 completed in 27.3 seconds with avg training loss 306.29\n",
      "epoch 417 completed in 26.45 seconds with avg training loss 306.2\n",
      "epoch 418 completed in 26.92 seconds with avg training loss 305.96\n",
      "epoch 419 completed in 26.91 seconds with avg training loss 306.21\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 420 completed in 26.3 seconds with avg training loss 306.07\n",
      "model not saved. validation loss was  803.785390625\n",
      "epoch 421 completed in 26.72 seconds with avg training loss 306.12\n",
      "epoch 422 completed in 27.14 seconds with avg training loss 306.13\n",
      "epoch 423 completed in 27.42 seconds with avg training loss 305.8\n",
      "epoch 424 completed in 26.7 seconds with avg training loss 306.06\n",
      "epoch 425 completed in 26.94 seconds with avg training loss 305.87\n",
      "model not saved. validation loss was  809.2802148437501\n",
      "epoch 426 completed in 26.56 seconds with avg training loss 306.35\n",
      "epoch 427 completed in 26.95 seconds with avg training loss 306.01\n",
      "epoch 428 completed in 27.11 seconds with avg training loss 305.98\n",
      "epoch 429 completed in 26.77 seconds with avg training loss 305.97\n",
      "epoch 430 completed in 26.38 seconds with avg training loss 306.13\n",
      "model not saved. validation loss was  811.58259765625\n",
      "epoch 431 completed in 26.63 seconds with avg training loss 305.99\n",
      "epoch 432 completed in 26.73 seconds with avg training loss 306.15\n",
      "epoch 433 completed in 26.55 seconds with avg training loss 306.07\n",
      "epoch 434 completed in 27.0 seconds with avg training loss 306.79\n",
      "epoch 435 completed in 27.31 seconds with avg training loss 306.55\n",
      "model not saved. validation loss was  806.94552734375\n",
      "epoch 436 completed in 27.0 seconds with avg training loss 306.2\n",
      "epoch 437 completed in 27.33 seconds with avg training loss 306.93\n",
      "epoch 438 completed in 27.17 seconds with avg training loss 306.24\n",
      "epoch 439 completed in 27.09 seconds with avg training loss 306.05\n",
      "epoch 440 completed in 27.36 seconds with avg training loss 305.81\n",
      "model not saved. validation loss was  807.59404296875\n",
      "epoch 441 completed in 26.84 seconds with avg training loss 305.75\n",
      "epoch 442 completed in 26.91 seconds with avg training loss 305.92\n",
      "epoch 443 completed in 27.09 seconds with avg training loss 306.0\n",
      "epoch 444 completed in 27.11 seconds with avg training loss 306.46\n",
      "epoch 445 completed in 27.38 seconds with avg training loss 306.63\n",
      "model not saved. validation loss was  805.3624218749999\n",
      "epoch 446 completed in 27.09 seconds with avg training loss 306.27\n",
      "epoch 447 completed in 27.31 seconds with avg training loss 306.4\n",
      "epoch 448 completed in 27.28 seconds with avg training loss 305.95\n",
      "epoch 449 completed in 27.22 seconds with avg training loss 306.59\n",
      "epoch 450 completed in 27.17 seconds with avg training loss 306.2\n",
      "model not saved. validation loss was  808.8422265625\n",
      "epoch 451 completed in 27.17 seconds with avg training loss 306.04\n",
      "epoch 452 completed in 26.67 seconds with avg training loss 306.54\n",
      "epoch 453 completed in 26.7 seconds with avg training loss 305.87\n",
      "epoch 454 completed in 27.0 seconds with avg training loss 306.51\n",
      "epoch 455 completed in 27.03 seconds with avg training loss 305.81\n",
      "model not saved. validation loss was  806.352578125\n",
      "epoch 456 completed in 26.67 seconds with avg training loss 305.75\n",
      "epoch 457 completed in 26.61 seconds with avg training loss 306.32\n",
      "epoch 458 completed in 26.78 seconds with avg training loss 306.49\n",
      "epoch 459 completed in 26.58 seconds with avg training loss 306.48\n",
      "epoch 460 completed in 27.33 seconds with avg training loss 306.4\n",
      "model not saved. validation loss was  805.1907812500001\n",
      "epoch 461 completed in 26.84 seconds with avg training loss 305.95\n",
      "epoch 462 completed in 26.73 seconds with avg training loss 306.4\n",
      "epoch 463 completed in 27.02 seconds with avg training loss 306.05\n",
      "epoch 464 completed in 26.88 seconds with avg training loss 305.85\n",
      "epoch 465 completed in 26.98 seconds with avg training loss 305.96\n",
      "model not saved. validation loss was  803.93142578125\n",
      "epoch 466 completed in 26.69 seconds with avg training loss 306.59\n",
      "epoch 467 completed in 26.52 seconds with avg training loss 306.04\n",
      "epoch 468 completed in 26.56 seconds with avg training loss 306.11\n",
      "epoch 469 completed in 27.09 seconds with avg training loss 306.34\n",
      "epoch 470 completed in 26.72 seconds with avg training loss 306.33\n",
      "model not saved. validation loss was  810.44796875\n",
      "epoch 471 completed in 26.89 seconds with avg training loss 306.34\n",
      "epoch 472 completed in 26.91 seconds with avg training loss 306.29\n",
      "epoch 473 completed in 27.2 seconds with avg training loss 306.1\n",
      "epoch 474 completed in 26.52 seconds with avg training loss 306.05\n",
      "epoch 475 completed in 26.84 seconds with avg training loss 306.44\n",
      "model not saved. validation loss was  806.4526953125\n",
      "epoch 476 completed in 26.98 seconds with avg training loss 306.58\n",
      "epoch 477 completed in 27.45 seconds with avg training loss 306.76\n",
      "epoch 478 completed in 27.38 seconds with avg training loss 306.19\n",
      "epoch 479 completed in 27.05 seconds with avg training loss 305.93\n",
      "epoch 480 completed in 27.27 seconds with avg training loss 306.29\n",
      "model not saved. validation loss was  809.7495703125\n",
      "epoch 481 completed in 26.8 seconds with avg training loss 306.03\n",
      "epoch 482 completed in 27.25 seconds with avg training loss 306.59\n",
      "epoch 483 completed in 27.3 seconds with avg training loss 306.36\n",
      "epoch 484 completed in 26.67 seconds with avg training loss 305.99\n",
      "epoch 485 completed in 26.63 seconds with avg training loss 306.45\n",
      "model not saved. validation loss was  806.91478515625\n",
      "epoch 486 completed in 26.3 seconds with avg training loss 306.41\n",
      "epoch 487 completed in 26.94 seconds with avg training loss 306.16\n",
      "epoch 488 completed in 26.97 seconds with avg training loss 306.57\n",
      "epoch 489 completed in 26.64 seconds with avg training loss 306.23\n",
      "epoch 490 completed in 27.0 seconds with avg training loss 306.33\n",
      "model not saved. validation loss was  805.17330078125\n",
      "epoch 491 completed in 27.0 seconds with avg training loss 306.33\n",
      "epoch 492 completed in 26.83 seconds with avg training loss 305.92\n",
      "epoch 493 completed in 26.69 seconds with avg training loss 305.92\n",
      "epoch 494 completed in 26.83 seconds with avg training loss 305.97\n",
      "epoch 495 completed in 26.53 seconds with avg training loss 306.2\n",
      "model not saved. validation loss was  809.153828125\n",
      "epoch 496 completed in 26.81 seconds with avg training loss 305.99\n",
      "epoch 497 completed in 26.63 seconds with avg training loss 306.17\n",
      "epoch 498 completed in 27.23 seconds with avg training loss 306.15\n",
      "epoch 499 completed in 26.94 seconds with avg training loss 306.34\n",
      "epoch 500 completed in 26.86 seconds with avg training loss 305.51\n",
      "model not saved. validation loss was  800.247265625\n",
      "epoch 501 completed in 26.83 seconds with avg training loss 306.42\n",
      "epoch 502 completed in 27.11 seconds with avg training loss 306.19\n",
      "epoch 503 completed in 27.02 seconds with avg training loss 305.94\n",
      "epoch 504 completed in 26.63 seconds with avg training loss 306.11\n",
      "epoch 505 completed in 27.16 seconds with avg training loss 305.97\n",
      "model not saved. validation loss was  810.6189843750001\n",
      "epoch 506 completed in 26.86 seconds with avg training loss 306.65\n",
      "epoch 507 completed in 26.55 seconds with avg training loss 306.17\n",
      "epoch 508 completed in 26.61 seconds with avg training loss 306.7\n",
      "epoch 509 completed in 26.73 seconds with avg training loss 306.07\n",
      "epoch 510 completed in 26.64 seconds with avg training loss 306.24\n",
      "model not saved. validation loss was  805.87044921875\n",
      "epoch 511 completed in 26.64 seconds with avg training loss 305.81\n",
      "epoch 512 completed in 27.13 seconds with avg training loss 306.06\n",
      "epoch 513 completed in 27.05 seconds with avg training loss 306.24\n",
      "epoch 514 completed in 26.95 seconds with avg training loss 305.8\n",
      "epoch 515 completed in 26.91 seconds with avg training loss 306.29\n",
      "model not saved. validation loss was  808.7244335937501\n",
      "epoch 516 completed in 26.56 seconds with avg training loss 305.88\n",
      "epoch 517 completed in 27.14 seconds with avg training loss 305.86\n",
      "epoch 518 completed in 26.27 seconds with avg training loss 305.95\n",
      "epoch 519 completed in 26.92 seconds with avg training loss 306.19\n",
      "epoch 520 completed in 27.02 seconds with avg training loss 306.13\n",
      "model not saved. validation loss was  808.33955078125\n",
      "epoch 521 completed in 26.61 seconds with avg training loss 305.98\n",
      "epoch 522 completed in 26.91 seconds with avg training loss 306.58\n",
      "epoch 523 completed in 26.61 seconds with avg training loss 306.01\n",
      "epoch 524 completed in 26.8 seconds with avg training loss 306.32\n",
      "epoch 525 completed in 27.05 seconds with avg training loss 306.56\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model not saved. validation loss was  806.6598828125001\n",
      "epoch 526 completed in 27.16 seconds with avg training loss 305.82\n",
      "epoch 527 completed in 27.63 seconds with avg training loss 305.64\n",
      "epoch 528 completed in 26.94 seconds with avg training loss 306.66\n",
      "epoch 529 completed in 26.86 seconds with avg training loss 306.25\n",
      "epoch 530 completed in 27.17 seconds with avg training loss 305.96\n",
      "model not saved. validation loss was  799.5964453125\n",
      "epoch 531 completed in 26.97 seconds with avg training loss 306.0\n",
      "epoch 532 completed in 27.22 seconds with avg training loss 306.16\n",
      "epoch 533 completed in 26.69 seconds with avg training loss 306.16\n",
      "epoch 534 completed in 26.5 seconds with avg training loss 306.15\n",
      "epoch 535 completed in 27.13 seconds with avg training loss 306.28\n",
      "model not saved. validation loss was  801.95578125\n",
      "epoch 536 completed in 26.48 seconds with avg training loss 306.54\n",
      "epoch 537 completed in 26.78 seconds with avg training loss 306.34\n",
      "epoch 538 completed in 27.13 seconds with avg training loss 306.32\n",
      "epoch 539 completed in 27.36 seconds with avg training loss 306.23\n",
      "epoch 540 completed in 27.39 seconds with avg training loss 305.65\n",
      "model not saved. validation loss was  802.070703125\n",
      "epoch 541 completed in 27.42 seconds with avg training loss 306.09\n",
      "epoch 542 completed in 27.31 seconds with avg training loss 305.78\n",
      "epoch 543 completed in 27.05 seconds with avg training loss 306.59\n",
      "epoch 544 completed in 26.72 seconds with avg training loss 306.35\n",
      "epoch 545 completed in 27.11 seconds with avg training loss 306.04\n",
      "model not saved. validation loss was  804.71873046875\n",
      "epoch 546 completed in 26.89 seconds with avg training loss 306.15\n",
      "epoch 547 completed in 27.08 seconds with avg training loss 305.52\n",
      "epoch 548 completed in 27.05 seconds with avg training loss 306.76\n",
      "epoch 549 completed in 27.11 seconds with avg training loss 306.5\n",
      "epoch 550 completed in 27.44 seconds with avg training loss 306.42\n",
      "model not saved. validation loss was  803.59515625\n",
      "epoch 551 completed in 27.05 seconds with avg training loss 306.43\n",
      "epoch 552 completed in 27.34 seconds with avg training loss 306.08\n",
      "epoch 553 completed in 26.8 seconds with avg training loss 306.42\n",
      "epoch 554 completed in 27.23 seconds with avg training loss 306.42\n",
      "epoch 555 completed in 26.7 seconds with avg training loss 306.91\n",
      "model not saved. validation loss was  802.7220703125\n",
      "epoch 556 completed in 26.77 seconds with avg training loss 305.98\n",
      "epoch 557 completed in 27.39 seconds with avg training loss 306.11\n",
      "epoch 558 completed in 27.03 seconds with avg training loss 306.35\n",
      "epoch 559 completed in 27.61 seconds with avg training loss 305.38\n",
      "epoch 560 completed in 27.44 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  803.3594726562501\n",
      "epoch 561 completed in 26.78 seconds with avg training loss 306.66\n",
      "epoch 562 completed in 27.19 seconds with avg training loss 306.28\n",
      "epoch 563 completed in 27.33 seconds with avg training loss 306.46\n",
      "epoch 564 completed in 27.0 seconds with avg training loss 306.38\n",
      "epoch 565 completed in 26.97 seconds with avg training loss 306.0\n",
      "model not saved. validation loss was  803.53666015625\n",
      "epoch 566 completed in 27.34 seconds with avg training loss 306.05\n",
      "epoch 567 completed in 27.33 seconds with avg training loss 306.65\n",
      "epoch 568 completed in 27.09 seconds with avg training loss 306.24\n",
      "epoch 569 completed in 26.81 seconds with avg training loss 305.84\n",
      "epoch 570 completed in 26.89 seconds with avg training loss 306.5\n",
      "model not saved. validation loss was  800.15650390625\n",
      "epoch 571 completed in 26.75 seconds with avg training loss 306.22\n",
      "epoch 572 completed in 26.45 seconds with avg training loss 306.47\n",
      "epoch 573 completed in 26.81 seconds with avg training loss 306.52\n",
      "epoch 574 completed in 26.78 seconds with avg training loss 306.11\n",
      "epoch 575 completed in 26.72 seconds with avg training loss 306.39\n",
      "model not saved. validation loss was  809.8937109374999\n",
      "epoch 576 completed in 26.61 seconds with avg training loss 306.18\n",
      "epoch 577 completed in 26.72 seconds with avg training loss 305.8\n",
      "epoch 578 completed in 26.73 seconds with avg training loss 305.91\n",
      "epoch 579 completed in 27.09 seconds with avg training loss 306.21\n",
      "epoch 580 completed in 26.73 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  799.5234960937499\n",
      "epoch 581 completed in 26.66 seconds with avg training loss 305.93\n",
      "epoch 582 completed in 26.97 seconds with avg training loss 305.82\n",
      "epoch 583 completed in 26.94 seconds with avg training loss 306.63\n",
      "epoch 584 completed in 26.75 seconds with avg training loss 305.64\n",
      "epoch 585 completed in 26.73 seconds with avg training loss 305.92\n",
      "model not saved. validation loss was  807.52544921875\n",
      "epoch 586 completed in 26.84 seconds with avg training loss 306.02\n",
      "epoch 587 completed in 26.64 seconds with avg training loss 306.61\n",
      "epoch 588 completed in 26.88 seconds with avg training loss 306.33\n",
      "epoch 589 completed in 27.23 seconds with avg training loss 305.61\n",
      "epoch 590 completed in 27.22 seconds with avg training loss 306.12\n",
      "model not saved. validation loss was  815.20462890625\n",
      "epoch 591 completed in 27.13 seconds with avg training loss 306.26\n",
      "epoch 592 completed in 27.31 seconds with avg training loss 306.44\n",
      "epoch 593 completed in 27.17 seconds with avg training loss 306.44\n",
      "epoch 594 completed in 27.13 seconds with avg training loss 306.2\n",
      "epoch 595 completed in 27.14 seconds with avg training loss 306.18\n",
      "model not saved. validation loss was  802.28060546875\n",
      "epoch 596 completed in 27.11 seconds with avg training loss 306.69\n",
      "epoch 597 completed in 27.06 seconds with avg training loss 306.5\n",
      "epoch 598 completed in 27.13 seconds with avg training loss 306.34\n",
      "epoch 599 completed in 26.89 seconds with avg training loss 306.19\n",
      "epoch 600 completed in 27.13 seconds with avg training loss 306.09\n",
      "model not saved. validation loss was  801.03341796875\n",
      "epoch 601 completed in 26.78 seconds with avg training loss 306.68\n",
      "epoch 602 completed in 26.48 seconds with avg training loss 305.72\n",
      "epoch 603 completed in 26.3 seconds with avg training loss 306.92\n",
      "epoch 604 completed in 26.41 seconds with avg training loss 306.18\n",
      "epoch 605 completed in 27.08 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  808.5460156250001\n",
      "epoch 606 completed in 27.08 seconds with avg training loss 306.03\n",
      "epoch 607 completed in 27.2 seconds with avg training loss 305.9\n",
      "epoch 608 completed in 27.06 seconds with avg training loss 306.28\n",
      "epoch 609 completed in 26.66 seconds with avg training loss 306.75\n",
      "epoch 610 completed in 26.84 seconds with avg training loss 305.99\n",
      "model not saved. validation loss was  803.76806640625\n",
      "epoch 611 completed in 26.92 seconds with avg training loss 305.74\n",
      "epoch 612 completed in 26.64 seconds with avg training loss 306.86\n",
      "epoch 613 completed in 26.58 seconds with avg training loss 306.02\n",
      "epoch 614 completed in 26.73 seconds with avg training loss 306.13\n",
      "epoch 615 completed in 26.94 seconds with avg training loss 306.7\n",
      "model not saved. validation loss was  807.9348828125\n",
      "epoch 616 completed in 26.66 seconds with avg training loss 306.39\n",
      "epoch 617 completed in 26.78 seconds with avg training loss 306.38\n",
      "epoch 618 completed in 26.73 seconds with avg training loss 305.7\n",
      "epoch 619 completed in 26.94 seconds with avg training loss 306.09\n",
      "epoch 620 completed in 26.7 seconds with avg training loss 306.48\n",
      "model not saved. validation loss was  808.863046875\n",
      "epoch 621 completed in 26.98 seconds with avg training loss 306.67\n",
      "epoch 622 completed in 27.58 seconds with avg training loss 306.22\n",
      "epoch 623 completed in 27.0 seconds with avg training loss 305.89\n",
      "epoch 624 completed in 27.19 seconds with avg training loss 306.33\n",
      "epoch 625 completed in 27.53 seconds with avg training loss 306.41\n",
      "model not saved. validation loss was  805.5741796875\n",
      "epoch 626 completed in 26.92 seconds with avg training loss 305.95\n",
      "epoch 627 completed in 27.13 seconds with avg training loss 306.59\n",
      "epoch 628 completed in 26.95 seconds with avg training loss 306.05\n",
      "epoch 629 completed in 27.22 seconds with avg training loss 306.4\n",
      "epoch 630 completed in 27.16 seconds with avg training loss 306.3\n",
      "model not saved. validation loss was  802.5462109375001\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 631 completed in 26.94 seconds with avg training loss 306.16\n",
      "epoch 632 completed in 27.19 seconds with avg training loss 306.48\n",
      "epoch 633 completed in 27.48 seconds with avg training loss 306.18\n",
      "epoch 634 completed in 26.97 seconds with avg training loss 306.29\n",
      "epoch 635 completed in 27.11 seconds with avg training loss 305.95\n",
      "model not saved. validation loss was  803.404765625\n",
      "epoch 636 completed in 26.88 seconds with avg training loss 306.38\n",
      "epoch 637 completed in 26.66 seconds with avg training loss 305.94\n",
      "epoch 638 completed in 27.03 seconds with avg training loss 305.97\n",
      "epoch 639 completed in 26.73 seconds with avg training loss 306.0\n",
      "epoch 640 completed in 26.44 seconds with avg training loss 306.33\n",
      "model not saved. validation loss was  807.27453125\n",
      "epoch 641 completed in 27.23 seconds with avg training loss 306.02\n",
      "epoch 642 completed in 26.88 seconds with avg training loss 305.94\n",
      "epoch 643 completed in 27.73 seconds with avg training loss 306.18\n",
      "epoch 644 completed in 27.11 seconds with avg training loss 305.89\n",
      "epoch 645 completed in 27.13 seconds with avg training loss 306.47\n",
      "model not saved. validation loss was  808.4490234375\n",
      "epoch 646 completed in 27.23 seconds with avg training loss 306.28\n",
      "epoch 647 completed in 27.33 seconds with avg training loss 306.29\n",
      "epoch 648 completed in 27.14 seconds with avg training loss 306.17\n",
      "epoch 649 completed in 27.14 seconds with avg training loss 306.22\n",
      "epoch 650 completed in 26.63 seconds with avg training loss 306.64\n",
      "model not saved. validation loss was  807.8315625\n",
      "epoch 651 completed in 27.06 seconds with avg training loss 306.6\n",
      "epoch 652 completed in 26.61 seconds with avg training loss 306.16\n",
      "epoch 653 completed in 26.94 seconds with avg training loss 305.68\n",
      "epoch 654 completed in 27.11 seconds with avg training loss 306.64\n",
      "epoch 655 completed in 27.27 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  809.22021484375\n",
      "epoch 656 completed in 26.88 seconds with avg training loss 306.19\n",
      "epoch 657 completed in 26.92 seconds with avg training loss 306.05\n",
      "epoch 658 completed in 26.91 seconds with avg training loss 306.07\n",
      "epoch 659 completed in 26.97 seconds with avg training loss 306.44\n",
      "epoch 660 completed in 27.06 seconds with avg training loss 306.37\n",
      "model not saved. validation loss was  801.8253125\n",
      "epoch 661 completed in 27.03 seconds with avg training loss 306.32\n",
      "epoch 662 completed in 27.25 seconds with avg training loss 306.18\n",
      "epoch 663 completed in 27.27 seconds with avg training loss 306.07\n",
      "epoch 664 completed in 27.2 seconds with avg training loss 306.24\n",
      "epoch 665 completed in 27.09 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  808.6949609375001\n",
      "epoch 666 completed in 27.02 seconds with avg training loss 305.82\n",
      "epoch 667 completed in 27.08 seconds with avg training loss 306.2\n",
      "epoch 668 completed in 27.31 seconds with avg training loss 305.72\n",
      "epoch 669 completed in 26.59 seconds with avg training loss 306.03\n",
      "epoch 670 completed in 27.27 seconds with avg training loss 306.47\n",
      "model not saved. validation loss was  803.3895507812499\n",
      "epoch 671 completed in 26.89 seconds with avg training loss 306.3\n",
      "epoch 672 completed in 26.89 seconds with avg training loss 306.45\n",
      "epoch 673 completed in 26.63 seconds with avg training loss 305.64\n",
      "epoch 674 completed in 27.11 seconds with avg training loss 305.95\n",
      "epoch 675 completed in 26.75 seconds with avg training loss 306.51\n",
      "model not saved. validation loss was  807.1272265625\n",
      "epoch 676 completed in 27.19 seconds with avg training loss 305.96\n",
      "epoch 677 completed in 27.11 seconds with avg training loss 306.18\n",
      "epoch 678 completed in 27.02 seconds with avg training loss 306.72\n",
      "epoch 679 completed in 26.94 seconds with avg training loss 306.43\n",
      "epoch 680 completed in 26.72 seconds with avg training loss 306.32\n",
      "model not saved. validation loss was  806.0829296874999\n",
      "epoch 681 completed in 26.7 seconds with avg training loss 305.71\n",
      "epoch 682 completed in 26.7 seconds with avg training loss 306.33\n",
      "epoch 683 completed in 26.45 seconds with avg training loss 306.47\n",
      "epoch 684 completed in 26.28 seconds with avg training loss 306.36\n",
      "epoch 685 completed in 27.06 seconds with avg training loss 306.35\n",
      "model not saved. validation loss was  807.68421875\n",
      "epoch 686 completed in 27.16 seconds with avg training loss 306.05\n",
      "epoch 687 completed in 26.78 seconds with avg training loss 306.2\n",
      "epoch 688 completed in 27.16 seconds with avg training loss 305.86\n",
      "epoch 689 completed in 26.73 seconds with avg training loss 305.93\n",
      "epoch 690 completed in 27.36 seconds with avg training loss 306.25\n",
      "model not saved. validation loss was  806.739296875\n",
      "epoch 691 completed in 26.66 seconds with avg training loss 306.21\n",
      "epoch 692 completed in 27.09 seconds with avg training loss 306.5\n",
      "epoch 693 completed in 26.91 seconds with avg training loss 305.98\n",
      "epoch 694 completed in 26.67 seconds with avg training loss 305.91\n",
      "epoch 695 completed in 27.13 seconds with avg training loss 306.4\n",
      "model not saved. validation loss was  802.93861328125\n",
      "epoch 696 completed in 26.42 seconds with avg training loss 305.72\n",
      "epoch 697 completed in 26.94 seconds with avg training loss 306.08\n",
      "epoch 698 completed in 26.38 seconds with avg training loss 306.25\n",
      "epoch 699 completed in 26.61 seconds with avg training loss 306.31\n",
      "epoch 700 completed in 26.81 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  808.01703125\n",
      "epoch 701 completed in 26.52 seconds with avg training loss 306.06\n",
      "epoch 702 completed in 26.83 seconds with avg training loss 305.63\n",
      "epoch 703 completed in 27.16 seconds with avg training loss 306.21\n",
      "epoch 704 completed in 27.34 seconds with avg training loss 306.31\n",
      "epoch 705 completed in 27.22 seconds with avg training loss 306.05\n",
      "model not saved. validation loss was  801.97119140625\n",
      "epoch 706 completed in 27.19 seconds with avg training loss 306.28\n",
      "epoch 707 completed in 27.3 seconds with avg training loss 306.12\n",
      "epoch 708 completed in 27.16 seconds with avg training loss 306.01\n",
      "epoch 709 completed in 27.09 seconds with avg training loss 306.45\n",
      "epoch 710 completed in 27.3 seconds with avg training loss 306.36\n",
      "model not saved. validation loss was  812.2583593749999\n",
      "epoch 711 completed in 27.33 seconds with avg training loss 306.75\n",
      "epoch 712 completed in 27.3 seconds with avg training loss 306.13\n",
      "epoch 713 completed in 27.06 seconds with avg training loss 306.69\n",
      "epoch 714 completed in 27.23 seconds with avg training loss 306.19\n",
      "epoch 715 completed in 27.52 seconds with avg training loss 306.27\n",
      "model not saved. validation loss was  804.95314453125\n",
      "epoch 716 completed in 26.89 seconds with avg training loss 306.27\n",
      "epoch 717 completed in 27.36 seconds with avg training loss 306.22\n",
      "epoch 718 completed in 27.36 seconds with avg training loss 306.25\n",
      "epoch 719 completed in 27.25 seconds with avg training loss 306.32\n",
      "epoch 720 completed in 27.23 seconds with avg training loss 306.6\n",
      "model not saved. validation loss was  804.4509960937501\n",
      "epoch 721 completed in 27.28 seconds with avg training loss 306.18\n",
      "epoch 722 completed in 27.39 seconds with avg training loss 305.86\n",
      "epoch 723 completed in 27.09 seconds with avg training loss 306.0\n",
      "epoch 724 completed in 27.27 seconds with avg training loss 306.45\n",
      "epoch 725 completed in 27.55 seconds with avg training loss 305.81\n",
      "model not saved. validation loss was  808.6953320312499\n",
      "epoch 726 completed in 27.09 seconds with avg training loss 306.22\n",
      "epoch 727 completed in 27.36 seconds with avg training loss 306.01\n",
      "epoch 728 completed in 27.45 seconds with avg training loss 306.31\n",
      "epoch 729 completed in 27.25 seconds with avg training loss 305.75\n",
      "epoch 730 completed in 27.3 seconds with avg training loss 306.56\n",
      "model not saved. validation loss was  810.98361328125\n",
      "epoch 731 completed in 26.89 seconds with avg training loss 306.33\n",
      "epoch 732 completed in 27.34 seconds with avg training loss 305.93\n",
      "epoch 733 completed in 27.27 seconds with avg training loss 306.02\n",
      "epoch 734 completed in 27.22 seconds with avg training loss 305.91\n",
      "epoch 735 completed in 27.23 seconds with avg training loss 305.7\n",
      "model not saved. validation loss was  803.08455078125\n",
      "epoch 736 completed in 27.31 seconds with avg training loss 306.6\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 737 completed in 27.23 seconds with avg training loss 305.77\n",
      "epoch 738 completed in 27.05 seconds with avg training loss 306.7\n",
      "epoch 739 completed in 27.27 seconds with avg training loss 305.71\n",
      "epoch 740 completed in 27.13 seconds with avg training loss 306.44\n",
      "model not saved. validation loss was  806.025078125\n",
      "epoch 741 completed in 27.25 seconds with avg training loss 306.36\n",
      "epoch 742 completed in 27.13 seconds with avg training loss 306.36\n",
      "epoch 743 completed in 27.0 seconds with avg training loss 306.45\n",
      "epoch 744 completed in 27.56 seconds with avg training loss 306.66\n",
      "epoch 745 completed in 27.22 seconds with avg training loss 306.13\n",
      "model not saved. validation loss was  806.17201171875\n",
      "epoch 746 completed in 27.22 seconds with avg training loss 306.53\n",
      "epoch 747 completed in 27.55 seconds with avg training loss 306.2\n",
      "epoch 748 completed in 27.34 seconds with avg training loss 306.27\n",
      "epoch 749 completed in 27.48 seconds with avg training loss 305.98\n",
      "epoch 750 completed in 27.5 seconds with avg training loss 305.88\n",
      "model not saved. validation loss was  809.067109375\n",
      "epoch 751 completed in 27.52 seconds with avg training loss 306.21\n",
      "epoch 752 completed in 27.47 seconds with avg training loss 305.95\n",
      "epoch 753 completed in 27.34 seconds with avg training loss 306.01\n",
      "epoch 754 completed in 27.14 seconds with avg training loss 306.16\n",
      "epoch 755 completed in 27.92 seconds with avg training loss 306.51\n",
      "model not saved. validation loss was  812.91533203125\n",
      "epoch 756 completed in 27.42 seconds with avg training loss 305.9\n",
      "epoch 757 completed in 27.55 seconds with avg training loss 306.15\n",
      "epoch 758 completed in 26.77 seconds with avg training loss 305.82\n",
      "epoch 759 completed in 27.45 seconds with avg training loss 306.27\n",
      "epoch 760 completed in 27.39 seconds with avg training loss 306.49\n",
      "model not saved. validation loss was  804.6716015625\n",
      "epoch 761 completed in 27.23 seconds with avg training loss 306.73\n",
      "epoch 762 completed in 27.48 seconds with avg training loss 306.19\n",
      "epoch 763 completed in 27.31 seconds with avg training loss 306.56\n",
      "epoch 764 completed in 27.53 seconds with avg training loss 306.08\n",
      "epoch 765 completed in 27.56 seconds with avg training loss 305.67\n",
      "model not saved. validation loss was  798.7325976562499\n",
      "epoch 766 completed in 26.89 seconds with avg training loss 306.3\n",
      "epoch 767 completed in 27.47 seconds with avg training loss 305.98\n",
      "epoch 768 completed in 27.3 seconds with avg training loss 306.32\n",
      "epoch 769 completed in 27.67 seconds with avg training loss 306.43\n",
      "epoch 770 completed in 27.25 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  801.7771875000001\n",
      "epoch 771 completed in 27.2 seconds with avg training loss 306.15\n",
      "epoch 772 completed in 27.89 seconds with avg training loss 305.74\n",
      "epoch 773 completed in 27.41 seconds with avg training loss 306.32\n",
      "epoch 774 completed in 27.42 seconds with avg training loss 306.02\n",
      "epoch 775 completed in 27.5 seconds with avg training loss 306.1\n",
      "model not saved. validation loss was  795.13529296875\n",
      "epoch 776 completed in 27.45 seconds with avg training loss 306.48\n",
      "epoch 777 completed in 27.19 seconds with avg training loss 306.09\n",
      "epoch 778 completed in 27.14 seconds with avg training loss 306.1\n",
      "epoch 779 completed in 27.06 seconds with avg training loss 306.49\n",
      "epoch 780 completed in 27.27 seconds with avg training loss 305.84\n",
      "model not saved. validation loss was  805.98220703125\n",
      "epoch 781 completed in 27.28 seconds with avg training loss 306.89\n",
      "epoch 782 completed in 27.3 seconds with avg training loss 306.35\n",
      "epoch 783 completed in 27.36 seconds with avg training loss 306.11\n",
      "epoch 784 completed in 26.95 seconds with avg training loss 306.72\n",
      "epoch 785 completed in 26.89 seconds with avg training loss 305.76\n",
      "model not saved. validation loss was  796.3765039062499\n",
      "epoch 786 completed in 27.34 seconds with avg training loss 306.49\n",
      "epoch 787 completed in 27.23 seconds with avg training loss 306.12\n",
      "epoch 788 completed in 27.45 seconds with avg training loss 305.98\n",
      "epoch 789 completed in 27.19 seconds with avg training loss 306.04\n",
      "epoch 790 completed in 27.75 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  808.54578125\n",
      "epoch 791 completed in 27.27 seconds with avg training loss 305.92\n",
      "epoch 792 completed in 27.72 seconds with avg training loss 305.71\n",
      "epoch 793 completed in 27.48 seconds with avg training loss 306.35\n",
      "epoch 794 completed in 27.36 seconds with avg training loss 306.31\n",
      "epoch 795 completed in 27.63 seconds with avg training loss 305.68\n",
      "model not saved. validation loss was  801.6996093750001\n",
      "epoch 796 completed in 26.47 seconds with avg training loss 305.92\n",
      "epoch 797 completed in 27.09 seconds with avg training loss 305.91\n",
      "epoch 798 completed in 27.42 seconds with avg training loss 306.23\n",
      "epoch 799 completed in 27.36 seconds with avg training loss 306.33\n",
      "epoch 800 completed in 26.92 seconds with avg training loss 306.94\n",
      "model not saved. validation loss was  805.57791015625\n",
      "epoch 801 completed in 27.52 seconds with avg training loss 306.71\n",
      "epoch 802 completed in 27.28 seconds with avg training loss 306.7\n",
      "epoch 803 completed in 27.28 seconds with avg training loss 306.49\n",
      "epoch 804 completed in 27.19 seconds with avg training loss 306.32\n",
      "epoch 805 completed in 27.14 seconds with avg training loss 305.68\n",
      "model not saved. validation loss was  803.05259765625\n",
      "epoch 806 completed in 27.23 seconds with avg training loss 305.81\n",
      "epoch 807 completed in 27.78 seconds with avg training loss 306.43\n",
      "epoch 808 completed in 27.53 seconds with avg training loss 306.11\n",
      "epoch 809 completed in 27.11 seconds with avg training loss 305.57\n",
      "epoch 810 completed in 27.02 seconds with avg training loss 306.64\n",
      "model not saved. validation loss was  806.1230859375\n",
      "epoch 811 completed in 27.63 seconds with avg training loss 306.12\n",
      "epoch 812 completed in 27.19 seconds with avg training loss 306.16\n",
      "epoch 813 completed in 27.27 seconds with avg training loss 306.0\n",
      "epoch 814 completed in 27.56 seconds with avg training loss 306.32\n",
      "epoch 815 completed in 27.31 seconds with avg training loss 306.16\n",
      "model not saved. validation loss was  806.36951171875\n",
      "epoch 816 completed in 27.06 seconds with avg training loss 305.98\n",
      "epoch 817 completed in 27.48 seconds with avg training loss 306.61\n",
      "epoch 818 completed in 27.53 seconds with avg training loss 305.95\n",
      "epoch 819 completed in 27.22 seconds with avg training loss 305.75\n",
      "epoch 820 completed in 27.36 seconds with avg training loss 305.9\n",
      "model not saved. validation loss was  805.50619140625\n",
      "epoch 821 completed in 27.5 seconds with avg training loss 306.39\n",
      "epoch 822 completed in 27.52 seconds with avg training loss 306.09\n",
      "epoch 823 completed in 27.34 seconds with avg training loss 306.7\n",
      "epoch 824 completed in 27.7 seconds with avg training loss 306.2\n",
      "epoch 825 completed in 27.66 seconds with avg training loss 306.05\n",
      "model not saved. validation loss was  806.0340039062501\n",
      "epoch 826 completed in 27.2 seconds with avg training loss 305.96\n",
      "epoch 827 completed in 26.91 seconds with avg training loss 306.56\n",
      "epoch 828 completed in 27.42 seconds with avg training loss 305.78\n",
      "epoch 829 completed in 27.42 seconds with avg training loss 306.18\n",
      "epoch 830 completed in 27.47 seconds with avg training loss 305.8\n",
      "model not saved. validation loss was  808.2528710937501\n",
      "epoch 831 completed in 26.83 seconds with avg training loss 306.48\n",
      "epoch 832 completed in 27.03 seconds with avg training loss 306.01\n",
      "epoch 833 completed in 27.89 seconds with avg training loss 306.83\n",
      "epoch 834 completed in 27.41 seconds with avg training loss 306.09\n",
      "epoch 835 completed in 27.31 seconds with avg training loss 306.16\n",
      "model not saved. validation loss was  812.31734375\n",
      "epoch 836 completed in 27.25 seconds with avg training loss 305.92\n",
      "epoch 837 completed in 26.95 seconds with avg training loss 306.5\n",
      "epoch 838 completed in 27.31 seconds with avg training loss 306.07\n",
      "epoch 839 completed in 26.94 seconds with avg training loss 306.18\n",
      "epoch 840 completed in 27.45 seconds with avg training loss 305.72\n",
      "model not saved. validation loss was  807.4425390624999\n",
      "epoch 841 completed in 27.17 seconds with avg training loss 306.62\n",
      "epoch 842 completed in 27.31 seconds with avg training loss 306.23\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 843 completed in 27.47 seconds with avg training loss 306.49\n",
      "epoch 844 completed in 27.38 seconds with avg training loss 306.47\n",
      "epoch 845 completed in 26.94 seconds with avg training loss 306.04\n",
      "model not saved. validation loss was  798.56166015625\n",
      "epoch 846 completed in 27.36 seconds with avg training loss 306.39\n",
      "epoch 847 completed in 26.98 seconds with avg training loss 305.99\n",
      "epoch 848 completed in 27.28 seconds with avg training loss 306.22\n",
      "epoch 849 completed in 27.0 seconds with avg training loss 306.1\n",
      "epoch 850 completed in 26.78 seconds with avg training loss 306.31\n",
      "model not saved. validation loss was  803.270859375\n",
      "epoch 851 completed in 27.09 seconds with avg training loss 306.63\n",
      "epoch 852 completed in 26.94 seconds with avg training loss 306.13\n",
      "epoch 853 completed in 26.91 seconds with avg training loss 306.29\n",
      "epoch 854 completed in 26.63 seconds with avg training loss 306.12\n",
      "epoch 855 completed in 26.98 seconds with avg training loss 305.97\n",
      "model not saved. validation loss was  806.3606835937501\n",
      "epoch 856 completed in 26.92 seconds with avg training loss 306.08\n",
      "epoch 857 completed in 26.92 seconds with avg training loss 305.97\n",
      "epoch 858 completed in 27.14 seconds with avg training loss 306.51\n",
      "epoch 859 completed in 27.05 seconds with avg training loss 306.52\n",
      "epoch 860 completed in 27.25 seconds with avg training loss 305.63\n",
      "model not saved. validation loss was  806.47365234375\n",
      "epoch 861 completed in 26.89 seconds with avg training loss 306.45\n",
      "epoch 862 completed in 26.94 seconds with avg training loss 305.87\n",
      "epoch 863 completed in 26.94 seconds with avg training loss 306.26\n",
      "epoch 864 completed in 27.39 seconds with avg training loss 305.98\n",
      "epoch 865 completed in 27.05 seconds with avg training loss 305.53\n",
      "model not saved. validation loss was  805.5536328124999\n",
      "epoch 866 completed in 26.83 seconds with avg training loss 305.8\n",
      "epoch 867 completed in 27.39 seconds with avg training loss 306.19\n",
      "epoch 868 completed in 27.06 seconds with avg training loss 306.15\n",
      "epoch 869 completed in 27.14 seconds with avg training loss 305.8\n",
      "epoch 870 completed in 27.08 seconds with avg training loss 306.59\n",
      "model not saved. validation loss was  802.8420703125\n",
      "epoch 871 completed in 27.19 seconds with avg training loss 306.67\n",
      "epoch 872 completed in 27.47 seconds with avg training loss 305.85\n",
      "epoch 873 completed in 27.19 seconds with avg training loss 306.09\n",
      "epoch 874 completed in 27.22 seconds with avg training loss 306.02\n",
      "epoch 875 completed in 26.7 seconds with avg training loss 305.71\n",
      "model not saved. validation loss was  801.2501757812499\n",
      "epoch 876 completed in 26.97 seconds with avg training loss 306.65\n",
      "epoch 877 completed in 27.11 seconds with avg training loss 306.42\n",
      "epoch 878 completed in 27.02 seconds with avg training loss 306.07\n",
      "epoch 879 completed in 26.97 seconds with avg training loss 305.94\n",
      "epoch 880 completed in 27.28 seconds with avg training loss 306.09\n",
      "model not saved. validation loss was  803.4148828125001\n",
      "epoch 881 completed in 26.7 seconds with avg training loss 306.4\n",
      "epoch 882 completed in 26.86 seconds with avg training loss 305.9\n",
      "epoch 883 completed in 27.31 seconds with avg training loss 305.98\n",
      "epoch 884 completed in 26.98 seconds with avg training loss 306.38\n",
      "epoch 885 completed in 27.09 seconds with avg training loss 306.12\n",
      "model not saved. validation loss was  812.54982421875\n",
      "epoch 886 completed in 27.14 seconds with avg training loss 305.73\n",
      "epoch 887 completed in 27.38 seconds with avg training loss 306.61\n",
      "epoch 888 completed in 27.23 seconds with avg training loss 306.25\n",
      "epoch 889 completed in 27.31 seconds with avg training loss 306.09\n",
      "epoch 890 completed in 26.78 seconds with avg training loss 306.45\n",
      "model not saved. validation loss was  805.6080078125001\n",
      "epoch 891 completed in 27.08 seconds with avg training loss 305.98\n",
      "epoch 892 completed in 27.36 seconds with avg training loss 306.54\n",
      "epoch 893 completed in 26.73 seconds with avg training loss 306.27\n",
      "epoch 894 completed in 27.17 seconds with avg training loss 306.22\n",
      "epoch 895 completed in 26.56 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  806.9029101562501\n",
      "epoch 896 completed in 26.56 seconds with avg training loss 306.35\n",
      "epoch 897 completed in 27.14 seconds with avg training loss 305.91\n",
      "epoch 898 completed in 27.11 seconds with avg training loss 306.46\n",
      "epoch 899 completed in 26.56 seconds with avg training loss 306.47\n",
      "epoch 900 completed in 27.23 seconds with avg training loss 305.58\n",
      "model not saved. validation loss was  803.05861328125\n",
      "epoch 901 completed in 26.92 seconds with avg training loss 306.0\n",
      "epoch 902 completed in 27.11 seconds with avg training loss 306.76\n",
      "epoch 903 completed in 26.86 seconds with avg training loss 305.58\n",
      "epoch 904 completed in 26.7 seconds with avg training loss 305.81\n",
      "epoch 905 completed in 26.61 seconds with avg training loss 306.21\n",
      "model not saved. validation loss was  807.33126953125\n",
      "epoch 906 completed in 27.27 seconds with avg training loss 306.48\n",
      "epoch 907 completed in 27.41 seconds with avg training loss 306.5\n",
      "epoch 908 completed in 27.03 seconds with avg training loss 306.22\n",
      "epoch 909 completed in 27.25 seconds with avg training loss 305.97\n",
      "epoch 910 completed in 26.73 seconds with avg training loss 305.99\n",
      "model not saved. validation loss was  809.04517578125\n",
      "epoch 911 completed in 27.2 seconds with avg training loss 306.37\n",
      "epoch 912 completed in 27.06 seconds with avg training loss 306.26\n",
      "epoch 913 completed in 26.73 seconds with avg training loss 306.45\n",
      "epoch 914 completed in 26.78 seconds with avg training loss 306.82\n",
      "epoch 915 completed in 27.08 seconds with avg training loss 306.43\n",
      "model not saved. validation loss was  806.5370703125\n",
      "epoch 916 completed in 27.08 seconds with avg training loss 306.23\n",
      "epoch 917 completed in 27.44 seconds with avg training loss 306.2\n",
      "epoch 918 completed in 27.17 seconds with avg training loss 306.08\n",
      "epoch 919 completed in 27.16 seconds with avg training loss 306.42\n",
      "epoch 920 completed in 27.28 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  806.60205078125\n",
      "epoch 921 completed in 26.86 seconds with avg training loss 306.1\n",
      "epoch 922 completed in 26.91 seconds with avg training loss 305.63\n",
      "epoch 923 completed in 27.23 seconds with avg training loss 306.29\n",
      "epoch 924 completed in 27.16 seconds with avg training loss 306.09\n",
      "epoch 925 completed in 27.56 seconds with avg training loss 305.93\n",
      "model not saved. validation loss was  805.78853515625\n",
      "epoch 926 completed in 27.19 seconds with avg training loss 305.95\n",
      "epoch 927 completed in 27.42 seconds with avg training loss 305.94\n",
      "epoch 928 completed in 27.3 seconds with avg training loss 306.33\n",
      "epoch 929 completed in 26.75 seconds with avg training loss 306.42\n",
      "epoch 930 completed in 27.02 seconds with avg training loss 306.05\n",
      "model not saved. validation loss was  805.8522070312499\n",
      "epoch 931 completed in 26.89 seconds with avg training loss 306.26\n",
      "epoch 932 completed in 27.11 seconds with avg training loss 306.15\n",
      "epoch 933 completed in 26.88 seconds with avg training loss 305.99\n",
      "epoch 934 completed in 27.31 seconds with avg training loss 306.51\n",
      "epoch 935 completed in 27.27 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  798.16705078125\n",
      "epoch 936 completed in 27.0 seconds with avg training loss 306.02\n",
      "epoch 937 completed in 26.64 seconds with avg training loss 306.42\n",
      "epoch 938 completed in 26.81 seconds with avg training loss 306.54\n",
      "epoch 939 completed in 26.84 seconds with avg training loss 306.94\n",
      "epoch 940 completed in 26.92 seconds with avg training loss 306.35\n",
      "model not saved. validation loss was  812.1865625\n",
      "epoch 941 completed in 26.72 seconds with avg training loss 306.62\n",
      "epoch 942 completed in 27.23 seconds with avg training loss 306.67\n",
      "epoch 943 completed in 26.64 seconds with avg training loss 306.41\n",
      "epoch 944 completed in 26.91 seconds with avg training loss 305.93\n",
      "epoch 945 completed in 27.3 seconds with avg training loss 305.61\n",
      "model not saved. validation loss was  797.32716796875\n",
      "epoch 946 completed in 27.02 seconds with avg training loss 306.36\n",
      "epoch 947 completed in 26.66 seconds with avg training loss 306.64\n",
      "epoch 948 completed in 27.27 seconds with avg training loss 306.16\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 949 completed in 27.2 seconds with avg training loss 305.79\n",
      "epoch 950 completed in 26.95 seconds with avg training loss 306.31\n",
      "model not saved. validation loss was  814.2908398437501\n",
      "epoch 951 completed in 27.17 seconds with avg training loss 305.88\n",
      "epoch 952 completed in 26.67 seconds with avg training loss 305.84\n",
      "epoch 953 completed in 27.08 seconds with avg training loss 306.17\n",
      "epoch 954 completed in 26.92 seconds with avg training loss 305.81\n",
      "epoch 955 completed in 27.36 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  803.058515625\n",
      "epoch 956 completed in 26.86 seconds with avg training loss 305.84\n",
      "epoch 957 completed in 27.0 seconds with avg training loss 306.15\n",
      "epoch 958 completed in 27.2 seconds with avg training loss 306.2\n",
      "epoch 959 completed in 26.59 seconds with avg training loss 306.15\n",
      "epoch 960 completed in 26.88 seconds with avg training loss 306.37\n",
      "model not saved. validation loss was  806.67001953125\n",
      "epoch 961 completed in 26.75 seconds with avg training loss 306.07\n",
      "epoch 962 completed in 27.06 seconds with avg training loss 306.02\n",
      "epoch 963 completed in 27.0 seconds with avg training loss 305.81\n",
      "epoch 964 completed in 27.06 seconds with avg training loss 306.63\n",
      "epoch 965 completed in 26.88 seconds with avg training loss 306.74\n",
      "model not saved. validation loss was  810.43078125\n",
      "epoch 966 completed in 27.05 seconds with avg training loss 305.88\n",
      "epoch 967 completed in 26.73 seconds with avg training loss 306.78\n",
      "epoch 968 completed in 26.88 seconds with avg training loss 306.44\n",
      "epoch 969 completed in 26.73 seconds with avg training loss 305.99\n",
      "epoch 970 completed in 27.06 seconds with avg training loss 306.43\n",
      "model not saved. validation loss was  808.67146484375\n",
      "epoch 971 completed in 26.81 seconds with avg training loss 306.22\n",
      "epoch 972 completed in 26.58 seconds with avg training loss 306.35\n",
      "epoch 973 completed in 27.36 seconds with avg training loss 306.12\n",
      "epoch 974 completed in 26.8 seconds with avg training loss 306.21\n",
      "epoch 975 completed in 27.11 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  808.1103515625\n",
      "epoch 976 completed in 26.56 seconds with avg training loss 306.23\n",
      "epoch 977 completed in 27.03 seconds with avg training loss 306.35\n",
      "epoch 978 completed in 27.19 seconds with avg training loss 306.25\n",
      "epoch 979 completed in 27.27 seconds with avg training loss 305.95\n",
      "epoch 980 completed in 27.03 seconds with avg training loss 305.91\n",
      "model not saved. validation loss was  804.33666015625\n",
      "epoch 981 completed in 27.13 seconds with avg training loss 306.46\n",
      "epoch 982 completed in 26.91 seconds with avg training loss 306.45\n",
      "epoch 983 completed in 26.7 seconds with avg training loss 306.22\n",
      "epoch 984 completed in 26.83 seconds with avg training loss 305.81\n",
      "epoch 985 completed in 27.03 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  809.8804882812499\n",
      "epoch 986 completed in 27.11 seconds with avg training loss 306.16\n",
      "epoch 987 completed in 27.06 seconds with avg training loss 306.27\n",
      "epoch 988 completed in 27.0 seconds with avg training loss 306.15\n",
      "epoch 989 completed in 26.97 seconds with avg training loss 306.22\n",
      "epoch 990 completed in 27.17 seconds with avg training loss 305.77\n",
      "model not saved. validation loss was  809.3953906250001\n",
      "epoch 991 completed in 26.69 seconds with avg training loss 306.03\n",
      "epoch 992 completed in 26.59 seconds with avg training loss 306.65\n",
      "epoch 993 completed in 27.14 seconds with avg training loss 306.44\n",
      "epoch 994 completed in 27.09 seconds with avg training loss 306.04\n",
      "epoch 995 completed in 26.86 seconds with avg training loss 306.76\n",
      "model not saved. validation loss was  806.82064453125\n",
      "epoch 996 completed in 26.98 seconds with avg training loss 306.36\n",
      "epoch 997 completed in 26.83 seconds with avg training loss 306.07\n",
      "epoch 998 completed in 26.56 seconds with avg training loss 306.51\n",
      "epoch 999 completed in 26.95 seconds with avg training loss 305.71\n",
      "epoch 1000 completed in 26.61 seconds with avg training loss 305.86\n",
      "model not saved. validation loss was  806.4753906249999\n",
      "epoch 1001 completed in 26.7 seconds with avg training loss 306.1\n",
      "epoch 1002 completed in 27.22 seconds with avg training loss 307.07\n",
      "epoch 1003 completed in 27.06 seconds with avg training loss 305.69\n",
      "epoch 1004 completed in 26.83 seconds with avg training loss 306.09\n",
      "epoch 1005 completed in 27.39 seconds with avg training loss 306.21\n",
      "model not saved. validation loss was  809.9887109375\n",
      "epoch 1006 completed in 27.25 seconds with avg training loss 306.3\n",
      "epoch 1007 completed in 27.19 seconds with avg training loss 305.9\n",
      "epoch 1008 completed in 26.94 seconds with avg training loss 306.06\n",
      "epoch 1009 completed in 26.73 seconds with avg training loss 306.4\n",
      "epoch 1010 completed in 26.97 seconds with avg training loss 306.35\n",
      "model not saved. validation loss was  802.2603320312501\n",
      "epoch 1011 completed in 26.88 seconds with avg training loss 306.59\n",
      "epoch 1012 completed in 27.11 seconds with avg training loss 306.22\n",
      "epoch 1013 completed in 26.86 seconds with avg training loss 305.98\n",
      "epoch 1014 completed in 26.89 seconds with avg training loss 306.04\n",
      "epoch 1015 completed in 27.27 seconds with avg training loss 306.38\n",
      "model not saved. validation loss was  798.66205078125\n",
      "epoch 1016 completed in 27.02 seconds with avg training loss 305.85\n",
      "epoch 1017 completed in 26.86 seconds with avg training loss 306.14\n",
      "epoch 1018 completed in 26.64 seconds with avg training loss 305.91\n",
      "epoch 1019 completed in 26.58 seconds with avg training loss 305.89\n",
      "epoch 1020 completed in 27.05 seconds with avg training loss 306.42\n",
      "model not saved. validation loss was  802.7513867187499\n",
      "epoch 1021 completed in 27.02 seconds with avg training loss 306.48\n",
      "epoch 1022 completed in 26.92 seconds with avg training loss 306.55\n",
      "epoch 1023 completed in 26.58 seconds with avg training loss 306.32\n",
      "epoch 1024 completed in 26.56 seconds with avg training loss 306.51\n",
      "epoch 1025 completed in 27.05 seconds with avg training loss 306.24\n",
      "model not saved. validation loss was  804.8463671874999\n",
      "epoch 1026 completed in 27.19 seconds with avg training loss 306.49\n",
      "epoch 1027 completed in 26.98 seconds with avg training loss 306.42\n",
      "epoch 1028 completed in 27.02 seconds with avg training loss 305.64\n",
      "epoch 1029 completed in 26.81 seconds with avg training loss 306.01\n",
      "epoch 1030 completed in 26.84 seconds with avg training loss 306.51\n",
      "model not saved. validation loss was  811.55345703125\n",
      "epoch 1031 completed in 26.84 seconds with avg training loss 305.9\n",
      "epoch 1032 completed in 26.97 seconds with avg training loss 306.4\n",
      "epoch 1033 completed in 26.83 seconds with avg training loss 306.35\n",
      "epoch 1034 completed in 26.86 seconds with avg training loss 306.55\n",
      "epoch 1035 completed in 26.91 seconds with avg training loss 306.07\n",
      "model not saved. validation loss was  809.37755859375\n",
      "epoch 1036 completed in 26.84 seconds with avg training loss 307.01\n",
      "epoch 1037 completed in 26.95 seconds with avg training loss 306.36\n",
      "epoch 1038 completed in 27.25 seconds with avg training loss 306.23\n",
      "epoch 1039 completed in 26.95 seconds with avg training loss 306.18\n",
      "epoch 1040 completed in 27.13 seconds with avg training loss 305.9\n",
      "model not saved. validation loss was  813.0732421875\n",
      "epoch 1041 completed in 26.81 seconds with avg training loss 306.14\n",
      "epoch 1042 completed in 27.14 seconds with avg training loss 306.31\n",
      "epoch 1043 completed in 26.36 seconds with avg training loss 306.56\n",
      "epoch 1044 completed in 26.94 seconds with avg training loss 306.19\n",
      "epoch 1045 completed in 27.14 seconds with avg training loss 306.24\n",
      "model not saved. validation loss was  805.7331835937499\n",
      "epoch 1046 completed in 27.17 seconds with avg training loss 306.15\n",
      "epoch 1047 completed in 27.22 seconds with avg training loss 306.43\n",
      "epoch 1048 completed in 26.95 seconds with avg training loss 306.56\n",
      "epoch 1049 completed in 26.8 seconds with avg training loss 305.88\n",
      "epoch 1050 completed in 26.7 seconds with avg training loss 306.53\n",
      "model not saved. validation loss was  811.92677734375\n",
      "epoch 1051 completed in 26.77 seconds with avg training loss 306.12\n",
      "epoch 1052 completed in 26.88 seconds with avg training loss 306.45\n",
      "epoch 1053 completed in 26.64 seconds with avg training loss 306.51\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1054 completed in 26.41 seconds with avg training loss 305.93\n",
      "epoch 1055 completed in 26.61 seconds with avg training loss 305.81\n",
      "model not saved. validation loss was  807.2160546875\n",
      "epoch 1056 completed in 26.94 seconds with avg training loss 306.21\n",
      "epoch 1057 completed in 26.48 seconds with avg training loss 305.99\n",
      "epoch 1058 completed in 26.88 seconds with avg training loss 306.21\n",
      "epoch 1059 completed in 27.03 seconds with avg training loss 306.21\n",
      "epoch 1060 completed in 26.77 seconds with avg training loss 306.4\n",
      "model not saved. validation loss was  803.4180078125\n",
      "epoch 1061 completed in 26.44 seconds with avg training loss 305.78\n",
      "epoch 1062 completed in 26.64 seconds with avg training loss 306.52\n",
      "epoch 1063 completed in 26.66 seconds with avg training loss 305.56\n",
      "epoch 1064 completed in 27.05 seconds with avg training loss 306.32\n",
      "epoch 1065 completed in 26.48 seconds with avg training loss 306.6\n",
      "model not saved. validation loss was  813.5914453124999\n",
      "epoch 1066 completed in 26.5 seconds with avg training loss 306.26\n",
      "epoch 1067 completed in 26.67 seconds with avg training loss 305.9\n",
      "epoch 1068 completed in 27.06 seconds with avg training loss 306.5\n",
      "epoch 1069 completed in 27.06 seconds with avg training loss 306.4\n",
      "epoch 1070 completed in 27.13 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  803.6354101562499\n",
      "epoch 1071 completed in 27.08 seconds with avg training loss 305.98\n",
      "epoch 1072 completed in 27.16 seconds with avg training loss 306.45\n",
      "epoch 1073 completed in 27.13 seconds with avg training loss 306.06\n",
      "epoch 1074 completed in 26.8 seconds with avg training loss 306.86\n",
      "epoch 1075 completed in 27.03 seconds with avg training loss 306.12\n",
      "model not saved. validation loss was  806.3587890625\n",
      "epoch 1076 completed in 26.39 seconds with avg training loss 306.76\n",
      "epoch 1077 completed in 26.66 seconds with avg training loss 306.38\n",
      "epoch 1078 completed in 26.56 seconds with avg training loss 306.39\n",
      "epoch 1079 completed in 27.28 seconds with avg training loss 305.94\n",
      "epoch 1080 completed in 26.86 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  800.3528515625\n",
      "epoch 1081 completed in 26.77 seconds with avg training loss 306.49\n",
      "epoch 1082 completed in 26.59 seconds with avg training loss 306.41\n",
      "epoch 1083 completed in 26.5 seconds with avg training loss 306.44\n",
      "epoch 1084 completed in 26.89 seconds with avg training loss 306.39\n",
      "epoch 1085 completed in 26.25 seconds with avg training loss 306.56\n",
      "model not saved. validation loss was  801.1874609375\n",
      "epoch 1086 completed in 26.88 seconds with avg training loss 306.24\n",
      "epoch 1087 completed in 27.25 seconds with avg training loss 306.08\n",
      "epoch 1088 completed in 26.94 seconds with avg training loss 306.25\n",
      "epoch 1089 completed in 26.95 seconds with avg training loss 305.83\n",
      "epoch 1090 completed in 26.89 seconds with avg training loss 305.43\n",
      "model not saved. validation loss was  798.5609374999999\n",
      "epoch 1091 completed in 26.72 seconds with avg training loss 306.53\n",
      "epoch 1092 completed in 26.38 seconds with avg training loss 306.03\n",
      "epoch 1093 completed in 26.45 seconds with avg training loss 305.72\n",
      "epoch 1094 completed in 27.0 seconds with avg training loss 306.15\n",
      "epoch 1095 completed in 26.77 seconds with avg training loss 305.89\n",
      "model not saved. validation loss was  801.6666796875\n",
      "epoch 1096 completed in 26.97 seconds with avg training loss 306.28\n",
      "epoch 1097 completed in 27.09 seconds with avg training loss 305.98\n",
      "epoch 1098 completed in 27.34 seconds with avg training loss 306.58\n",
      "epoch 1099 completed in 27.14 seconds with avg training loss 306.5\n",
      "epoch 1100 completed in 26.77 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  808.34654296875\n",
      "epoch 1101 completed in 26.83 seconds with avg training loss 306.01\n",
      "epoch 1102 completed in 27.31 seconds with avg training loss 306.53\n",
      "epoch 1103 completed in 27.16 seconds with avg training loss 307.02\n",
      "epoch 1104 completed in 27.23 seconds with avg training loss 305.72\n",
      "epoch 1105 completed in 27.5 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  802.9654101562501\n",
      "epoch 1106 completed in 26.94 seconds with avg training loss 306.33\n",
      "epoch 1107 completed in 26.89 seconds with avg training loss 306.43\n",
      "epoch 1108 completed in 26.7 seconds with avg training loss 306.17\n",
      "epoch 1109 completed in 26.45 seconds with avg training loss 305.95\n",
      "epoch 1110 completed in 27.09 seconds with avg training loss 306.45\n",
      "model not saved. validation loss was  798.52357421875\n",
      "epoch 1111 completed in 26.94 seconds with avg training loss 306.08\n",
      "epoch 1112 completed in 26.83 seconds with avg training loss 306.28\n",
      "epoch 1113 completed in 27.22 seconds with avg training loss 306.17\n",
      "epoch 1114 completed in 26.92 seconds with avg training loss 306.09\n",
      "epoch 1115 completed in 26.77 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  807.7132812499999\n",
      "epoch 1116 completed in 26.97 seconds with avg training loss 306.15\n",
      "epoch 1117 completed in 27.25 seconds with avg training loss 305.93\n",
      "epoch 1118 completed in 26.98 seconds with avg training loss 306.5\n",
      "epoch 1119 completed in 26.92 seconds with avg training loss 306.32\n",
      "epoch 1120 completed in 27.14 seconds with avg training loss 305.93\n",
      "model not saved. validation loss was  805.18873046875\n",
      "epoch 1121 completed in 27.2 seconds with avg training loss 306.14\n",
      "epoch 1122 completed in 27.31 seconds with avg training loss 306.39\n",
      "epoch 1123 completed in 27.03 seconds with avg training loss 306.18\n",
      "epoch 1124 completed in 27.09 seconds with avg training loss 306.39\n",
      "epoch 1125 completed in 26.84 seconds with avg training loss 306.35\n",
      "model not saved. validation loss was  805.6370703125\n",
      "epoch 1126 completed in 27.33 seconds with avg training loss 306.03\n",
      "epoch 1127 completed in 26.5 seconds with avg training loss 305.98\n",
      "epoch 1128 completed in 26.03 seconds with avg training loss 306.15\n",
      "epoch 1129 completed in 26.53 seconds with avg training loss 306.05\n",
      "epoch 1130 completed in 26.73 seconds with avg training loss 306.47\n",
      "model not saved. validation loss was  809.28275390625\n",
      "epoch 1131 completed in 26.88 seconds with avg training loss 306.06\n",
      "epoch 1132 completed in 27.02 seconds with avg training loss 305.96\n",
      "epoch 1133 completed in 26.95 seconds with avg training loss 306.63\n",
      "epoch 1134 completed in 26.78 seconds with avg training loss 305.89\n",
      "epoch 1135 completed in 27.14 seconds with avg training loss 306.18\n",
      "model not saved. validation loss was  807.76298828125\n",
      "epoch 1136 completed in 26.64 seconds with avg training loss 305.65\n",
      "epoch 1137 completed in 27.03 seconds with avg training loss 306.46\n",
      "epoch 1138 completed in 27.0 seconds with avg training loss 306.07\n",
      "epoch 1139 completed in 27.36 seconds with avg training loss 306.04\n",
      "epoch 1140 completed in 26.95 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  807.08095703125\n",
      "epoch 1141 completed in 27.27 seconds with avg training loss 306.47\n",
      "epoch 1142 completed in 27.08 seconds with avg training loss 306.3\n",
      "epoch 1143 completed in 27.08 seconds with avg training loss 306.15\n",
      "epoch 1144 completed in 26.81 seconds with avg training loss 306.28\n",
      "epoch 1145 completed in 26.53 seconds with avg training loss 306.24\n",
      "model not saved. validation loss was  806.72908203125\n",
      "epoch 1146 completed in 27.25 seconds with avg training loss 306.13\n",
      "epoch 1147 completed in 26.59 seconds with avg training loss 306.11\n",
      "epoch 1148 completed in 26.84 seconds with avg training loss 305.96\n",
      "epoch 1149 completed in 26.91 seconds with avg training loss 306.44\n",
      "epoch 1150 completed in 27.11 seconds with avg training loss 305.94\n",
      "model not saved. validation loss was  802.467265625\n",
      "epoch 1151 completed in 27.23 seconds with avg training loss 306.01\n",
      "epoch 1152 completed in 26.88 seconds with avg training loss 305.84\n",
      "epoch 1153 completed in 27.16 seconds with avg training loss 306.5\n",
      "epoch 1154 completed in 26.92 seconds with avg training loss 306.62\n",
      "epoch 1155 completed in 27.17 seconds with avg training loss 306.55\n",
      "model not saved. validation loss was  805.95998046875\n",
      "epoch 1156 completed in 26.94 seconds with avg training loss 306.63\n",
      "epoch 1157 completed in 27.31 seconds with avg training loss 306.27\n",
      "epoch 1158 completed in 27.52 seconds with avg training loss 305.8\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1159 completed in 27.23 seconds with avg training loss 306.2\n",
      "epoch 1160 completed in 27.42 seconds with avg training loss 306.21\n",
      "model not saved. validation loss was  803.311953125\n",
      "epoch 1161 completed in 27.08 seconds with avg training loss 306.15\n",
      "epoch 1162 completed in 27.0 seconds with avg training loss 305.98\n",
      "epoch 1163 completed in 26.97 seconds with avg training loss 306.1\n",
      "epoch 1164 completed in 27.3 seconds with avg training loss 306.2\n",
      "epoch 1165 completed in 27.05 seconds with avg training loss 306.58\n",
      "model not saved. validation loss was  806.32\n",
      "epoch 1166 completed in 27.34 seconds with avg training loss 306.76\n",
      "epoch 1167 completed in 27.45 seconds with avg training loss 306.01\n",
      "epoch 1168 completed in 27.13 seconds with avg training loss 306.12\n",
      "epoch 1169 completed in 26.97 seconds with avg training loss 306.34\n",
      "epoch 1170 completed in 26.98 seconds with avg training loss 306.45\n",
      "model not saved. validation loss was  807.6836328125\n",
      "epoch 1171 completed in 27.06 seconds with avg training loss 306.41\n",
      "epoch 1172 completed in 26.52 seconds with avg training loss 306.39\n",
      "epoch 1173 completed in 27.03 seconds with avg training loss 306.42\n",
      "epoch 1174 completed in 27.14 seconds with avg training loss 307.06\n",
      "epoch 1175 completed in 26.95 seconds with avg training loss 306.73\n",
      "model not saved. validation loss was  805.61232421875\n",
      "epoch 1176 completed in 26.73 seconds with avg training loss 306.0\n",
      "epoch 1177 completed in 27.02 seconds with avg training loss 306.83\n",
      "epoch 1178 completed in 26.95 seconds with avg training loss 305.78\n",
      "epoch 1179 completed in 27.05 seconds with avg training loss 305.87\n",
      "epoch 1180 completed in 27.19 seconds with avg training loss 306.57\n",
      "model not saved. validation loss was  809.899765625\n",
      "epoch 1181 completed in 27.38 seconds with avg training loss 306.34\n",
      "epoch 1182 completed in 27.22 seconds with avg training loss 306.19\n",
      "epoch 1183 completed in 26.78 seconds with avg training loss 306.49\n",
      "epoch 1184 completed in 27.06 seconds with avg training loss 306.12\n",
      "epoch 1185 completed in 26.86 seconds with avg training loss 306.64\n",
      "model not saved. validation loss was  809.8849218749999\n",
      "epoch 1186 completed in 27.13 seconds with avg training loss 306.42\n",
      "epoch 1187 completed in 26.67 seconds with avg training loss 306.5\n",
      "epoch 1188 completed in 27.53 seconds with avg training loss 306.07\n",
      "epoch 1189 completed in 26.97 seconds with avg training loss 305.63\n",
      "epoch 1190 completed in 26.75 seconds with avg training loss 306.43\n",
      "model not saved. validation loss was  806.14107421875\n",
      "epoch 1191 completed in 26.94 seconds with avg training loss 306.11\n",
      "epoch 1192 completed in 26.78 seconds with avg training loss 306.56\n",
      "epoch 1193 completed in 27.28 seconds with avg training loss 306.08\n",
      "epoch 1194 completed in 26.97 seconds with avg training loss 305.89\n",
      "epoch 1195 completed in 26.78 seconds with avg training loss 306.45\n",
      "model not saved. validation loss was  815.678984375\n",
      "epoch 1196 completed in 27.31 seconds with avg training loss 306.25\n",
      "epoch 1197 completed in 26.91 seconds with avg training loss 306.38\n",
      "epoch 1198 completed in 26.64 seconds with avg training loss 306.22\n",
      "epoch 1199 completed in 26.59 seconds with avg training loss 306.32\n",
      "epoch 1200 completed in 26.34 seconds with avg training loss 306.32\n",
      "model not saved. validation loss was  806.67998046875\n",
      "epoch 1201 completed in 26.55 seconds with avg training loss 305.99\n",
      "epoch 1202 completed in 27.0 seconds with avg training loss 306.25\n",
      "epoch 1203 completed in 27.28 seconds with avg training loss 306.23\n",
      "epoch 1204 completed in 27.13 seconds with avg training loss 306.52\n",
      "epoch 1205 completed in 27.03 seconds with avg training loss 306.01\n",
      "model not saved. validation loss was  809.9430078124999\n",
      "epoch 1206 completed in 27.03 seconds with avg training loss 306.43\n",
      "epoch 1207 completed in 27.03 seconds with avg training loss 305.74\n",
      "epoch 1208 completed in 26.97 seconds with avg training loss 305.97\n",
      "epoch 1209 completed in 27.55 seconds with avg training loss 306.18\n",
      "epoch 1210 completed in 27.47 seconds with avg training loss 306.24\n",
      "model not saved. validation loss was  801.241640625\n",
      "epoch 1211 completed in 27.06 seconds with avg training loss 305.97\n",
      "epoch 1212 completed in 27.02 seconds with avg training loss 305.92\n",
      "epoch 1213 completed in 27.63 seconds with avg training loss 306.06\n",
      "epoch 1214 completed in 27.27 seconds with avg training loss 305.85\n",
      "epoch 1215 completed in 27.45 seconds with avg training loss 306.46\n",
      "model not saved. validation loss was  811.55099609375\n",
      "epoch 1216 completed in 27.48 seconds with avg training loss 305.63\n",
      "epoch 1217 completed in 27.47 seconds with avg training loss 306.29\n",
      "epoch 1218 completed in 26.97 seconds with avg training loss 306.37\n",
      "epoch 1219 completed in 27.3 seconds with avg training loss 306.3\n",
      "epoch 1220 completed in 27.02 seconds with avg training loss 306.0\n",
      "model not saved. validation loss was  804.8559765625\n",
      "epoch 1221 completed in 27.08 seconds with avg training loss 306.24\n",
      "epoch 1222 completed in 27.11 seconds with avg training loss 306.64\n",
      "epoch 1223 completed in 27.09 seconds with avg training loss 306.14\n",
      "epoch 1224 completed in 27.28 seconds with avg training loss 306.67\n",
      "epoch 1225 completed in 27.42 seconds with avg training loss 306.07\n",
      "model not saved. validation loss was  804.68919921875\n",
      "epoch 1226 completed in 27.42 seconds with avg training loss 306.17\n",
      "epoch 1227 completed in 27.08 seconds with avg training loss 306.0\n",
      "epoch 1228 completed in 26.81 seconds with avg training loss 306.22\n",
      "epoch 1229 completed in 27.36 seconds with avg training loss 306.24\n",
      "epoch 1230 completed in 27.19 seconds with avg training loss 306.0\n",
      "model not saved. validation loss was  813.26595703125\n",
      "epoch 1231 completed in 27.56 seconds with avg training loss 306.07\n",
      "epoch 1232 completed in 27.33 seconds with avg training loss 306.47\n",
      "epoch 1233 completed in 27.55 seconds with avg training loss 305.74\n",
      "epoch 1234 completed in 27.3 seconds with avg training loss 306.26\n",
      "epoch 1235 completed in 27.42 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  805.442578125\n",
      "epoch 1236 completed in 26.95 seconds with avg training loss 306.22\n",
      "epoch 1237 completed in 27.16 seconds with avg training loss 305.72\n",
      "epoch 1238 completed in 26.95 seconds with avg training loss 306.7\n",
      "epoch 1239 completed in 27.48 seconds with avg training loss 306.01\n",
      "epoch 1240 completed in 27.22 seconds with avg training loss 305.6\n",
      "model not saved. validation loss was  804.3157421875001\n",
      "epoch 1241 completed in 27.0 seconds with avg training loss 306.38\n",
      "epoch 1242 completed in 27.09 seconds with avg training loss 306.69\n",
      "epoch 1243 completed in 26.34 seconds with avg training loss 306.34\n",
      "epoch 1244 completed in 26.61 seconds with avg training loss 306.34\n",
      "epoch 1245 completed in 26.5 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  809.84005859375\n",
      "epoch 1246 completed in 26.52 seconds with avg training loss 306.3\n",
      "epoch 1247 completed in 26.77 seconds with avg training loss 306.01\n",
      "epoch 1248 completed in 27.25 seconds with avg training loss 306.53\n",
      "epoch 1249 completed in 27.34 seconds with avg training loss 306.33\n",
      "epoch 1250 completed in 26.97 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  804.21861328125\n",
      "epoch 1251 completed in 27.2 seconds with avg training loss 306.72\n",
      "epoch 1252 completed in 27.02 seconds with avg training loss 305.64\n",
      "epoch 1253 completed in 27.22 seconds with avg training loss 306.43\n",
      "epoch 1254 completed in 26.88 seconds with avg training loss 305.96\n",
      "epoch 1255 completed in 27.23 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  807.7930664062501\n",
      "epoch 1256 completed in 27.34 seconds with avg training loss 306.13\n",
      "epoch 1257 completed in 26.81 seconds with avg training loss 306.37\n",
      "epoch 1258 completed in 27.09 seconds with avg training loss 305.72\n",
      "epoch 1259 completed in 26.95 seconds with avg training loss 306.12\n",
      "epoch 1260 completed in 26.77 seconds with avg training loss 305.88\n",
      "model not saved. validation loss was  803.04640625\n",
      "epoch 1261 completed in 27.31 seconds with avg training loss 306.13\n",
      "epoch 1262 completed in 27.06 seconds with avg training loss 306.39\n",
      "epoch 1263 completed in 27.0 seconds with avg training loss 306.42\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1264 completed in 26.75 seconds with avg training loss 305.67\n",
      "epoch 1265 completed in 27.5 seconds with avg training loss 306.49\n",
      "model not saved. validation loss was  806.99177734375\n",
      "epoch 1266 completed in 27.58 seconds with avg training loss 306.01\n",
      "epoch 1267 completed in 27.16 seconds with avg training loss 306.14\n",
      "epoch 1268 completed in 27.41 seconds with avg training loss 306.25\n",
      "epoch 1269 completed in 26.97 seconds with avg training loss 305.97\n",
      "epoch 1270 completed in 26.67 seconds with avg training loss 306.09\n",
      "model not saved. validation loss was  813.73603515625\n",
      "epoch 1271 completed in 27.03 seconds with avg training loss 306.37\n",
      "epoch 1272 completed in 26.75 seconds with avg training loss 306.41\n",
      "epoch 1273 completed in 27.0 seconds with avg training loss 306.13\n",
      "epoch 1274 completed in 27.0 seconds with avg training loss 306.43\n",
      "epoch 1275 completed in 26.34 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  809.6546679687501\n",
      "epoch 1276 completed in 27.16 seconds with avg training loss 306.68\n",
      "epoch 1277 completed in 27.19 seconds with avg training loss 306.29\n",
      "epoch 1278 completed in 27.25 seconds with avg training loss 306.06\n",
      "epoch 1279 completed in 27.03 seconds with avg training loss 306.53\n",
      "epoch 1280 completed in 27.0 seconds with avg training loss 306.44\n",
      "model not saved. validation loss was  802.3370117187501\n",
      "epoch 1281 completed in 27.22 seconds with avg training loss 306.22\n",
      "epoch 1282 completed in 26.8 seconds with avg training loss 306.58\n",
      "epoch 1283 completed in 27.0 seconds with avg training loss 306.44\n",
      "epoch 1284 completed in 27.13 seconds with avg training loss 306.33\n",
      "epoch 1285 completed in 26.94 seconds with avg training loss 306.28\n",
      "model not saved. validation loss was  808.27724609375\n",
      "epoch 1286 completed in 26.89 seconds with avg training loss 306.22\n",
      "epoch 1287 completed in 27.05 seconds with avg training loss 306.08\n",
      "epoch 1288 completed in 26.92 seconds with avg training loss 305.66\n",
      "epoch 1289 completed in 26.94 seconds with avg training loss 306.23\n",
      "epoch 1290 completed in 27.06 seconds with avg training loss 306.3\n",
      "model not saved. validation loss was  805.3375390625\n",
      "epoch 1291 completed in 26.97 seconds with avg training loss 306.5\n",
      "epoch 1292 completed in 26.91 seconds with avg training loss 306.01\n",
      "epoch 1293 completed in 27.39 seconds with avg training loss 306.04\n",
      "epoch 1294 completed in 26.98 seconds with avg training loss 306.48\n",
      "epoch 1295 completed in 27.14 seconds with avg training loss 306.18\n",
      "model not saved. validation loss was  810.979921875\n",
      "epoch 1296 completed in 26.97 seconds with avg training loss 306.09\n",
      "epoch 1297 completed in 26.81 seconds with avg training loss 305.69\n",
      "epoch 1298 completed in 27.0 seconds with avg training loss 306.19\n",
      "epoch 1299 completed in 26.53 seconds with avg training loss 306.02\n",
      "epoch 1300 completed in 27.2 seconds with avg training loss 306.37\n",
      "model not saved. validation loss was  800.6953515625\n",
      "epoch 1301 completed in 27.22 seconds with avg training loss 305.78\n",
      "epoch 1302 completed in 27.06 seconds with avg training loss 306.24\n",
      "epoch 1303 completed in 27.05 seconds with avg training loss 306.29\n",
      "epoch 1304 completed in 27.28 seconds with avg training loss 305.88\n",
      "epoch 1305 completed in 27.3 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  811.15810546875\n",
      "epoch 1306 completed in 27.25 seconds with avg training loss 306.49\n",
      "epoch 1307 completed in 26.94 seconds with avg training loss 305.93\n",
      "epoch 1308 completed in 27.44 seconds with avg training loss 306.28\n",
      "epoch 1309 completed in 27.34 seconds with avg training loss 306.05\n",
      "epoch 1310 completed in 27.31 seconds with avg training loss 306.28\n",
      "model not saved. validation loss was  810.4541015625\n",
      "epoch 1311 completed in 27.22 seconds with avg training loss 306.32\n",
      "epoch 1312 completed in 27.42 seconds with avg training loss 306.42\n",
      "epoch 1313 completed in 27.27 seconds with avg training loss 306.48\n",
      "epoch 1314 completed in 27.06 seconds with avg training loss 306.16\n",
      "epoch 1315 completed in 27.2 seconds with avg training loss 305.99\n",
      "model not saved. validation loss was  806.010546875\n",
      "epoch 1316 completed in 26.91 seconds with avg training loss 306.25\n",
      "epoch 1317 completed in 27.52 seconds with avg training loss 306.21\n",
      "epoch 1318 completed in 26.7 seconds with avg training loss 306.45\n",
      "epoch 1319 completed in 26.84 seconds with avg training loss 306.34\n",
      "epoch 1320 completed in 26.88 seconds with avg training loss 306.02\n",
      "model not saved. validation loss was  810.33111328125\n",
      "epoch 1321 completed in 27.02 seconds with avg training loss 306.3\n",
      "epoch 1322 completed in 27.25 seconds with avg training loss 305.77\n",
      "epoch 1323 completed in 26.86 seconds with avg training loss 306.31\n",
      "epoch 1324 completed in 27.0 seconds with avg training loss 306.05\n",
      "epoch 1325 completed in 27.14 seconds with avg training loss 306.07\n",
      "model not saved. validation loss was  799.36978515625\n",
      "epoch 1326 completed in 27.31 seconds with avg training loss 306.07\n",
      "epoch 1327 completed in 26.98 seconds with avg training loss 306.83\n",
      "epoch 1328 completed in 26.81 seconds with avg training loss 306.77\n",
      "epoch 1329 completed in 26.97 seconds with avg training loss 306.41\n",
      "epoch 1330 completed in 26.92 seconds with avg training loss 305.63\n",
      "model not saved. validation loss was  799.8697265625\n",
      "epoch 1331 completed in 27.03 seconds with avg training loss 305.92\n",
      "epoch 1332 completed in 26.88 seconds with avg training loss 306.47\n",
      "epoch 1333 completed in 27.27 seconds with avg training loss 306.23\n",
      "epoch 1334 completed in 27.33 seconds with avg training loss 306.6\n",
      "epoch 1335 completed in 26.94 seconds with avg training loss 305.97\n",
      "model not saved. validation loss was  804.11865234375\n",
      "epoch 1336 completed in 26.91 seconds with avg training loss 306.34\n",
      "epoch 1337 completed in 27.13 seconds with avg training loss 305.91\n",
      "epoch 1338 completed in 27.33 seconds with avg training loss 306.14\n",
      "epoch 1339 completed in 27.39 seconds with avg training loss 306.02\n",
      "epoch 1340 completed in 27.41 seconds with avg training loss 305.98\n",
      "model not saved. validation loss was  804.3951953125\n",
      "epoch 1341 completed in 27.27 seconds with avg training loss 306.48\n",
      "epoch 1342 completed in 26.91 seconds with avg training loss 305.95\n",
      "epoch 1343 completed in 27.25 seconds with avg training loss 306.4\n",
      "epoch 1344 completed in 27.09 seconds with avg training loss 305.53\n",
      "epoch 1345 completed in 27.25 seconds with avg training loss 306.91\n",
      "model not saved. validation loss was  804.63619140625\n",
      "epoch 1346 completed in 27.0 seconds with avg training loss 306.09\n",
      "epoch 1347 completed in 27.03 seconds with avg training loss 306.12\n",
      "epoch 1348 completed in 27.17 seconds with avg training loss 306.02\n",
      "epoch 1349 completed in 27.0 seconds with avg training loss 306.24\n",
      "epoch 1350 completed in 27.09 seconds with avg training loss 306.56\n",
      "model not saved. validation loss was  802.8638476562501\n",
      "epoch 1351 completed in 27.27 seconds with avg training loss 306.3\n",
      "epoch 1352 completed in 27.08 seconds with avg training loss 305.97\n",
      "epoch 1353 completed in 27.17 seconds with avg training loss 306.24\n",
      "epoch 1354 completed in 27.22 seconds with avg training loss 306.54\n",
      "epoch 1355 completed in 26.86 seconds with avg training loss 305.91\n",
      "model not saved. validation loss was  809.4876367187501\n",
      "epoch 1356 completed in 27.42 seconds with avg training loss 306.14\n",
      "epoch 1357 completed in 27.72 seconds with avg training loss 305.9\n",
      "epoch 1358 completed in 27.13 seconds with avg training loss 306.09\n",
      "epoch 1359 completed in 27.39 seconds with avg training loss 306.38\n",
      "epoch 1360 completed in 27.66 seconds with avg training loss 305.77\n",
      "model not saved. validation loss was  804.44896484375\n",
      "epoch 1361 completed in 27.42 seconds with avg training loss 306.17\n",
      "epoch 1362 completed in 27.33 seconds with avg training loss 305.73\n",
      "epoch 1363 completed in 27.11 seconds with avg training loss 306.3\n",
      "epoch 1364 completed in 26.97 seconds with avg training loss 306.14\n",
      "epoch 1365 completed in 27.34 seconds with avg training loss 306.58\n",
      "model not saved. validation loss was  805.731015625\n",
      "epoch 1366 completed in 27.36 seconds with avg training loss 305.97\n",
      "epoch 1367 completed in 27.45 seconds with avg training loss 306.06\n",
      "epoch 1368 completed in 27.67 seconds with avg training loss 306.48\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1369 completed in 27.47 seconds with avg training loss 305.69\n",
      "epoch 1370 completed in 27.78 seconds with avg training loss 305.83\n",
      "model not saved. validation loss was  805.2157421874999\n",
      "epoch 1371 completed in 27.77 seconds with avg training loss 306.27\n",
      "epoch 1372 completed in 27.08 seconds with avg training loss 306.14\n",
      "epoch 1373 completed in 26.89 seconds with avg training loss 306.0\n",
      "epoch 1374 completed in 27.56 seconds with avg training loss 306.14\n",
      "epoch 1375 completed in 27.33 seconds with avg training loss 305.93\n",
      "model not saved. validation loss was  800.6002734375\n",
      "epoch 1376 completed in 27.03 seconds with avg training loss 305.92\n",
      "epoch 1377 completed in 27.48 seconds with avg training loss 306.9\n",
      "epoch 1378 completed in 27.44 seconds with avg training loss 306.33\n",
      "epoch 1379 completed in 27.53 seconds with avg training loss 306.17\n",
      "epoch 1380 completed in 27.11 seconds with avg training loss 306.63\n",
      "model not saved. validation loss was  802.606171875\n",
      "epoch 1381 completed in 27.13 seconds with avg training loss 306.2\n",
      "epoch 1382 completed in 27.48 seconds with avg training loss 305.68\n",
      "epoch 1383 completed in 26.89 seconds with avg training loss 306.17\n",
      "epoch 1384 completed in 27.2 seconds with avg training loss 306.88\n",
      "epoch 1385 completed in 26.75 seconds with avg training loss 306.01\n",
      "model not saved. validation loss was  800.88552734375\n",
      "epoch 1386 completed in 27.0 seconds with avg training loss 305.96\n",
      "epoch 1387 completed in 27.23 seconds with avg training loss 306.03\n",
      "epoch 1388 completed in 26.81 seconds with avg training loss 306.06\n",
      "epoch 1389 completed in 27.09 seconds with avg training loss 306.3\n",
      "epoch 1390 completed in 27.17 seconds with avg training loss 306.31\n",
      "model not saved. validation loss was  808.1002539062499\n",
      "epoch 1391 completed in 27.34 seconds with avg training loss 306.15\n",
      "epoch 1392 completed in 27.09 seconds with avg training loss 306.52\n",
      "epoch 1393 completed in 27.17 seconds with avg training loss 306.37\n",
      "epoch 1394 completed in 27.25 seconds with avg training loss 306.3\n",
      "epoch 1395 completed in 27.19 seconds with avg training loss 306.28\n",
      "model not saved. validation loss was  802.2737695312501\n",
      "epoch 1396 completed in 27.28 seconds with avg training loss 305.84\n",
      "epoch 1397 completed in 27.28 seconds with avg training loss 306.09\n",
      "epoch 1398 completed in 27.45 seconds with avg training loss 306.0\n",
      "epoch 1399 completed in 27.09 seconds with avg training loss 306.43\n",
      "epoch 1400 completed in 27.48 seconds with avg training loss 306.44\n",
      "model not saved. validation loss was  801.89119140625\n",
      "epoch 1401 completed in 26.92 seconds with avg training loss 306.02\n",
      "epoch 1402 completed in 27.33 seconds with avg training loss 305.82\n",
      "epoch 1403 completed in 27.44 seconds with avg training loss 306.13\n",
      "epoch 1404 completed in 27.3 seconds with avg training loss 306.12\n",
      "epoch 1405 completed in 27.42 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  807.23591796875\n",
      "epoch 1406 completed in 27.19 seconds with avg training loss 306.35\n",
      "epoch 1407 completed in 27.61 seconds with avg training loss 306.08\n",
      "epoch 1408 completed in 27.55 seconds with avg training loss 306.02\n",
      "epoch 1409 completed in 27.52 seconds with avg training loss 306.44\n",
      "epoch 1410 completed in 27.36 seconds with avg training loss 305.85\n",
      "model not saved. validation loss was  810.65578125\n",
      "epoch 1411 completed in 26.86 seconds with avg training loss 306.18\n",
      "epoch 1412 completed in 27.47 seconds with avg training loss 306.06\n",
      "epoch 1413 completed in 27.25 seconds with avg training loss 306.22\n",
      "epoch 1414 completed in 27.78 seconds with avg training loss 306.49\n",
      "epoch 1415 completed in 27.45 seconds with avg training loss 305.72\n",
      "model not saved. validation loss was  804.6623828125\n",
      "epoch 1416 completed in 27.63 seconds with avg training loss 306.61\n",
      "epoch 1417 completed in 27.08 seconds with avg training loss 305.89\n",
      "epoch 1418 completed in 27.67 seconds with avg training loss 306.24\n",
      "epoch 1419 completed in 27.44 seconds with avg training loss 306.16\n",
      "epoch 1420 completed in 27.17 seconds with avg training loss 306.44\n",
      "model not saved. validation loss was  806.8529492187499\n",
      "epoch 1421 completed in 27.64 seconds with avg training loss 305.94\n",
      "epoch 1422 completed in 27.55 seconds with avg training loss 305.68\n",
      "epoch 1423 completed in 27.64 seconds with avg training loss 306.37\n",
      "epoch 1424 completed in 27.56 seconds with avg training loss 306.23\n",
      "epoch 1425 completed in 27.16 seconds with avg training loss 306.83\n",
      "model not saved. validation loss was  799.0099414062499\n",
      "epoch 1426 completed in 27.56 seconds with avg training loss 306.09\n",
      "epoch 1427 completed in 27.19 seconds with avg training loss 305.99\n",
      "epoch 1428 completed in 27.73 seconds with avg training loss 306.23\n",
      "epoch 1429 completed in 27.25 seconds with avg training loss 305.67\n",
      "epoch 1430 completed in 27.39 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  796.6413671875\n",
      "epoch 1431 completed in 27.48 seconds with avg training loss 306.33\n",
      "epoch 1432 completed in 27.08 seconds with avg training loss 306.19\n",
      "epoch 1433 completed in 27.3 seconds with avg training loss 306.16\n",
      "epoch 1434 completed in 27.39 seconds with avg training loss 306.28\n",
      "epoch 1435 completed in 27.13 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  799.90966796875\n",
      "epoch 1436 completed in 27.23 seconds with avg training loss 306.08\n",
      "epoch 1437 completed in 27.31 seconds with avg training loss 306.5\n",
      "epoch 1438 completed in 27.16 seconds with avg training loss 306.31\n",
      "epoch 1439 completed in 26.95 seconds with avg training loss 306.22\n",
      "epoch 1440 completed in 27.45 seconds with avg training loss 306.2\n",
      "model not saved. validation loss was  803.7319921875\n",
      "epoch 1441 completed in 27.36 seconds with avg training loss 306.31\n",
      "epoch 1442 completed in 27.14 seconds with avg training loss 306.91\n",
      "epoch 1443 completed in 27.63 seconds with avg training loss 305.97\n",
      "epoch 1444 completed in 27.7 seconds with avg training loss 306.58\n",
      "epoch 1445 completed in 27.59 seconds with avg training loss 306.16\n",
      "model not saved. validation loss was  807.2416796875\n",
      "epoch 1446 completed in 27.23 seconds with avg training loss 306.58\n",
      "epoch 1447 completed in 27.06 seconds with avg training loss 306.48\n",
      "epoch 1448 completed in 27.48 seconds with avg training loss 306.25\n",
      "epoch 1449 completed in 26.92 seconds with avg training loss 306.08\n",
      "epoch 1450 completed in 27.33 seconds with avg training loss 305.79\n",
      "model not saved. validation loss was  803.80125\n",
      "epoch 1451 completed in 27.36 seconds with avg training loss 306.63\n",
      "epoch 1452 completed in 27.53 seconds with avg training loss 305.98\n",
      "epoch 1453 completed in 27.19 seconds with avg training loss 306.47\n",
      "epoch 1454 completed in 28.06 seconds with avg training loss 305.97\n",
      "epoch 1455 completed in 27.09 seconds with avg training loss 306.12\n",
      "model not saved. validation loss was  811.19923828125\n",
      "epoch 1456 completed in 27.66 seconds with avg training loss 306.18\n",
      "epoch 1457 completed in 27.66 seconds with avg training loss 306.09\n",
      "epoch 1458 completed in 27.55 seconds with avg training loss 306.15\n",
      "epoch 1459 completed in 27.47 seconds with avg training loss 306.09\n",
      "epoch 1460 completed in 27.53 seconds with avg training loss 306.01\n",
      "model not saved. validation loss was  807.6553125\n",
      "epoch 1461 completed in 27.25 seconds with avg training loss 306.49\n",
      "epoch 1462 completed in 27.03 seconds with avg training loss 306.1\n",
      "epoch 1463 completed in 27.33 seconds with avg training loss 306.12\n",
      "epoch 1464 completed in 27.16 seconds with avg training loss 305.74\n",
      "epoch 1465 completed in 27.3 seconds with avg training loss 306.4\n",
      "model not saved. validation loss was  801.57525390625\n",
      "epoch 1466 completed in 27.34 seconds with avg training loss 305.46\n",
      "epoch 1467 completed in 27.33 seconds with avg training loss 306.67\n",
      "epoch 1468 completed in 27.7 seconds with avg training loss 305.71\n",
      "epoch 1469 completed in 27.41 seconds with avg training loss 306.44\n",
      "epoch 1470 completed in 27.81 seconds with avg training loss 306.27\n",
      "model not saved. validation loss was  804.0848046875\n",
      "epoch 1471 completed in 27.19 seconds with avg training loss 305.98\n",
      "epoch 1472 completed in 27.27 seconds with avg training loss 306.22\n",
      "epoch 1473 completed in 27.02 seconds with avg training loss 306.05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1474 completed in 27.05 seconds with avg training loss 306.18\n",
      "epoch 1475 completed in 26.8 seconds with avg training loss 306.61\n",
      "model not saved. validation loss was  801.798125\n",
      "epoch 1476 completed in 27.38 seconds with avg training loss 305.89\n",
      "epoch 1477 completed in 27.08 seconds with avg training loss 305.91\n",
      "epoch 1478 completed in 27.19 seconds with avg training loss 306.63\n",
      "epoch 1479 completed in 27.31 seconds with avg training loss 306.12\n",
      "epoch 1480 completed in 27.06 seconds with avg training loss 306.61\n",
      "model not saved. validation loss was  798.5910351562501\n",
      "epoch 1481 completed in 27.3 seconds with avg training loss 305.59\n",
      "epoch 1482 completed in 27.28 seconds with avg training loss 306.12\n",
      "epoch 1483 completed in 27.19 seconds with avg training loss 306.6\n",
      "epoch 1484 completed in 26.86 seconds with avg training loss 306.44\n",
      "epoch 1485 completed in 27.33 seconds with avg training loss 306.11\n",
      "model not saved. validation loss was  808.4225195312499\n",
      "epoch 1486 completed in 26.56 seconds with avg training loss 305.97\n",
      "epoch 1487 completed in 26.88 seconds with avg training loss 306.48\n",
      "epoch 1488 completed in 27.23 seconds with avg training loss 306.09\n",
      "epoch 1489 completed in 27.19 seconds with avg training loss 306.12\n",
      "epoch 1490 completed in 26.83 seconds with avg training loss 306.25\n",
      "model not saved. validation loss was  800.5021875\n",
      "epoch 1491 completed in 27.08 seconds with avg training loss 305.76\n",
      "epoch 1492 completed in 27.33 seconds with avg training loss 306.39\n",
      "epoch 1493 completed in 26.73 seconds with avg training loss 305.97\n",
      "epoch 1494 completed in 27.22 seconds with avg training loss 306.64\n",
      "epoch 1495 completed in 27.77 seconds with avg training loss 305.79\n",
      "model not saved. validation loss was  807.09892578125\n",
      "epoch 1496 completed in 26.45 seconds with avg training loss 305.79\n",
      "epoch 1497 completed in 27.23 seconds with avg training loss 306.01\n",
      "epoch 1498 completed in 26.97 seconds with avg training loss 305.86\n",
      "epoch 1499 completed in 27.14 seconds with avg training loss 306.25\n",
      "epoch 1500 completed in 27.25 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  806.10505859375\n",
      "epoch 1501 completed in 27.23 seconds with avg training loss 306.32\n",
      "epoch 1502 completed in 27.06 seconds with avg training loss 305.55\n",
      "epoch 1503 completed in 27.05 seconds with avg training loss 306.32\n",
      "epoch 1504 completed in 27.03 seconds with avg training loss 306.4\n",
      "epoch 1505 completed in 26.7 seconds with avg training loss 305.56\n",
      "model not saved. validation loss was  809.5040624999999\n",
      "epoch 1506 completed in 27.42 seconds with avg training loss 306.19\n",
      "epoch 1507 completed in 26.86 seconds with avg training loss 306.29\n",
      "epoch 1508 completed in 27.05 seconds with avg training loss 305.97\n",
      "epoch 1509 completed in 26.83 seconds with avg training loss 305.86\n",
      "epoch 1510 completed in 27.36 seconds with avg training loss 306.3\n",
      "model not saved. validation loss was  806.68880859375\n",
      "epoch 1511 completed in 27.25 seconds with avg training loss 306.35\n",
      "epoch 1512 completed in 27.06 seconds with avg training loss 306.13\n",
      "epoch 1513 completed in 27.73 seconds with avg training loss 306.56\n",
      "epoch 1514 completed in 26.95 seconds with avg training loss 305.98\n",
      "epoch 1515 completed in 27.3 seconds with avg training loss 306.44\n",
      "model not saved. validation loss was  808.5575000000001\n",
      "epoch 1516 completed in 27.09 seconds with avg training loss 305.76\n",
      "epoch 1517 completed in 27.06 seconds with avg training loss 306.09\n",
      "epoch 1518 completed in 27.61 seconds with avg training loss 306.53\n",
      "epoch 1519 completed in 27.2 seconds with avg training loss 306.05\n",
      "epoch 1520 completed in 26.91 seconds with avg training loss 306.46\n",
      "model not saved. validation loss was  807.04443359375\n",
      "epoch 1521 completed in 27.23 seconds with avg training loss 306.24\n",
      "epoch 1522 completed in 26.81 seconds with avg training loss 305.83\n",
      "epoch 1523 completed in 26.86 seconds with avg training loss 305.97\n",
      "epoch 1524 completed in 27.28 seconds with avg training loss 306.5\n",
      "epoch 1525 completed in 27.05 seconds with avg training loss 306.01\n",
      "model not saved. validation loss was  804.08888671875\n",
      "epoch 1526 completed in 26.95 seconds with avg training loss 306.11\n",
      "epoch 1527 completed in 26.88 seconds with avg training loss 305.79\n",
      "epoch 1528 completed in 26.84 seconds with avg training loss 305.97\n",
      "epoch 1529 completed in 27.42 seconds with avg training loss 306.17\n",
      "epoch 1530 completed in 27.13 seconds with avg training loss 305.98\n",
      "model not saved. validation loss was  803.12576171875\n",
      "epoch 1531 completed in 27.02 seconds with avg training loss 305.9\n",
      "epoch 1532 completed in 27.05 seconds with avg training loss 305.95\n",
      "epoch 1533 completed in 26.94 seconds with avg training loss 306.11\n",
      "epoch 1534 completed in 27.14 seconds with avg training loss 305.9\n",
      "epoch 1535 completed in 26.92 seconds with avg training loss 306.41\n",
      "model not saved. validation loss was  806.54560546875\n",
      "epoch 1536 completed in 27.28 seconds with avg training loss 306.34\n",
      "epoch 1537 completed in 27.33 seconds with avg training loss 306.27\n",
      "epoch 1538 completed in 27.52 seconds with avg training loss 306.09\n",
      "epoch 1539 completed in 27.48 seconds with avg training loss 306.36\n",
      "epoch 1540 completed in 27.3 seconds with avg training loss 305.89\n",
      "model not saved. validation loss was  809.4200976562499\n",
      "epoch 1541 completed in 27.58 seconds with avg training loss 306.19\n",
      "epoch 1542 completed in 27.36 seconds with avg training loss 306.05\n",
      "epoch 1543 completed in 27.56 seconds with avg training loss 306.12\n",
      "epoch 1544 completed in 27.44 seconds with avg training loss 305.9\n",
      "epoch 1545 completed in 27.36 seconds with avg training loss 305.42\n",
      "model not saved. validation loss was  802.5935546875\n",
      "epoch 1546 completed in 27.61 seconds with avg training loss 305.92\n",
      "epoch 1547 completed in 27.19 seconds with avg training loss 306.19\n",
      "epoch 1548 completed in 27.22 seconds with avg training loss 306.32\n",
      "epoch 1549 completed in 26.55 seconds with avg training loss 306.76\n",
      "epoch 1550 completed in 26.56 seconds with avg training loss 306.77\n",
      "model not saved. validation loss was  804.7062304687499\n",
      "epoch 1551 completed in 26.59 seconds with avg training loss 305.99\n",
      "epoch 1552 completed in 26.7 seconds with avg training loss 305.89\n",
      "epoch 1553 completed in 27.22 seconds with avg training loss 306.31\n",
      "epoch 1554 completed in 26.97 seconds with avg training loss 305.83\n",
      "epoch 1555 completed in 26.77 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  811.74271484375\n",
      "epoch 1556 completed in 27.27 seconds with avg training loss 306.28\n",
      "epoch 1557 completed in 27.16 seconds with avg training loss 306.18\n",
      "epoch 1558 completed in 27.47 seconds with avg training loss 306.06\n",
      "epoch 1559 completed in 26.95 seconds with avg training loss 306.36\n",
      "epoch 1560 completed in 26.77 seconds with avg training loss 305.58\n",
      "model not saved. validation loss was  802.568125\n",
      "epoch 1561 completed in 26.89 seconds with avg training loss 306.29\n",
      "epoch 1562 completed in 26.67 seconds with avg training loss 306.01\n",
      "epoch 1563 completed in 27.33 seconds with avg training loss 305.99\n",
      "epoch 1564 completed in 27.17 seconds with avg training loss 305.86\n",
      "epoch 1565 completed in 27.23 seconds with avg training loss 306.07\n",
      "model not saved. validation loss was  803.91951171875\n",
      "epoch 1566 completed in 27.33 seconds with avg training loss 306.18\n",
      "epoch 1567 completed in 27.48 seconds with avg training loss 306.09\n",
      "epoch 1568 completed in 27.48 seconds with avg training loss 306.23\n",
      "epoch 1569 completed in 27.14 seconds with avg training loss 306.26\n",
      "epoch 1570 completed in 27.27 seconds with avg training loss 305.67\n",
      "model not saved. validation loss was  809.01119140625\n",
      "epoch 1571 completed in 27.5 seconds with avg training loss 306.21\n",
      "epoch 1572 completed in 26.97 seconds with avg training loss 306.31\n",
      "epoch 1573 completed in 27.27 seconds with avg training loss 306.35\n",
      "epoch 1574 completed in 27.42 seconds with avg training loss 306.61\n",
      "epoch 1575 completed in 26.97 seconds with avg training loss 306.54\n",
      "model not saved. validation loss was  809.13244140625\n",
      "epoch 1576 completed in 27.7 seconds with avg training loss 306.63\n",
      "epoch 1577 completed in 27.19 seconds with avg training loss 306.13\n",
      "epoch 1578 completed in 27.56 seconds with avg training loss 306.84\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1579 completed in 27.59 seconds with avg training loss 306.48\n",
      "epoch 1580 completed in 27.73 seconds with avg training loss 306.84\n",
      "model not saved. validation loss was  802.8599218749999\n",
      "epoch 1581 completed in 27.16 seconds with avg training loss 306.4\n",
      "epoch 1582 completed in 26.97 seconds with avg training loss 306.27\n",
      "epoch 1583 completed in 27.55 seconds with avg training loss 306.23\n",
      "epoch 1584 completed in 27.25 seconds with avg training loss 306.23\n",
      "epoch 1585 completed in 27.78 seconds with avg training loss 306.57\n",
      "model not saved. validation loss was  804.7760156249999\n",
      "epoch 1586 completed in 27.5 seconds with avg training loss 305.95\n",
      "epoch 1587 completed in 27.22 seconds with avg training loss 306.33\n",
      "epoch 1588 completed in 27.48 seconds with avg training loss 306.41\n",
      "epoch 1589 completed in 26.48 seconds with avg training loss 306.46\n",
      "epoch 1590 completed in 27.03 seconds with avg training loss 305.94\n",
      "model not saved. validation loss was  809.7665625\n",
      "epoch 1591 completed in 26.89 seconds with avg training loss 306.16\n",
      "epoch 1592 completed in 27.03 seconds with avg training loss 306.0\n",
      "epoch 1593 completed in 27.09 seconds with avg training loss 306.0\n",
      "epoch 1594 completed in 27.03 seconds with avg training loss 306.54\n",
      "epoch 1595 completed in 27.11 seconds with avg training loss 305.8\n",
      "model not saved. validation loss was  805.85130859375\n",
      "epoch 1596 completed in 27.25 seconds with avg training loss 306.55\n",
      "epoch 1597 completed in 27.3 seconds with avg training loss 305.96\n",
      "epoch 1598 completed in 27.47 seconds with avg training loss 306.27\n",
      "epoch 1599 completed in 27.42 seconds with avg training loss 305.98\n",
      "epoch 1600 completed in 27.19 seconds with avg training loss 305.47\n",
      "model not saved. validation loss was  805.6558789062499\n",
      "epoch 1601 completed in 26.91 seconds with avg training loss 305.77\n",
      "epoch 1602 completed in 27.11 seconds with avg training loss 305.9\n",
      "epoch 1603 completed in 26.48 seconds with avg training loss 306.13\n",
      "epoch 1604 completed in 26.92 seconds with avg training loss 306.67\n",
      "epoch 1605 completed in 26.58 seconds with avg training loss 306.04\n",
      "model not saved. validation loss was  803.7976757812501\n",
      "epoch 1606 completed in 27.11 seconds with avg training loss 306.42\n",
      "epoch 1607 completed in 26.89 seconds with avg training loss 305.8\n",
      "epoch 1608 completed in 26.52 seconds with avg training loss 306.57\n",
      "epoch 1609 completed in 26.81 seconds with avg training loss 306.14\n",
      "epoch 1610 completed in 26.8 seconds with avg training loss 306.41\n",
      "model not saved. validation loss was  803.0308203124999\n",
      "epoch 1611 completed in 27.31 seconds with avg training loss 306.03\n",
      "epoch 1612 completed in 27.23 seconds with avg training loss 305.62\n",
      "epoch 1613 completed in 27.36 seconds with avg training loss 306.45\n",
      "epoch 1614 completed in 27.7 seconds with avg training loss 305.92\n",
      "epoch 1615 completed in 27.31 seconds with avg training loss 306.01\n",
      "model not saved. validation loss was  801.8055859374999\n",
      "epoch 1616 completed in 27.31 seconds with avg training loss 306.47\n",
      "epoch 1617 completed in 27.19 seconds with avg training loss 305.66\n",
      "epoch 1618 completed in 27.5 seconds with avg training loss 305.96\n",
      "epoch 1619 completed in 27.39 seconds with avg training loss 305.61\n",
      "epoch 1620 completed in 27.36 seconds with avg training loss 306.42\n",
      "model not saved. validation loss was  798.5085546875\n",
      "epoch 1621 completed in 27.14 seconds with avg training loss 306.12\n",
      "epoch 1622 completed in 27.31 seconds with avg training loss 306.74\n",
      "epoch 1623 completed in 27.05 seconds with avg training loss 305.82\n",
      "epoch 1624 completed in 27.33 seconds with avg training loss 305.82\n",
      "epoch 1625 completed in 26.78 seconds with avg training loss 306.23\n",
      "model not saved. validation loss was  804.52033203125\n",
      "epoch 1626 completed in 26.41 seconds with avg training loss 306.6\n",
      "epoch 1627 completed in 26.7 seconds with avg training loss 306.25\n",
      "epoch 1628 completed in 27.0 seconds with avg training loss 305.71\n",
      "epoch 1629 completed in 27.06 seconds with avg training loss 306.42\n",
      "epoch 1630 completed in 27.03 seconds with avg training loss 306.5\n",
      "model not saved. validation loss was  805.41337890625\n",
      "epoch 1631 completed in 27.17 seconds with avg training loss 306.08\n",
      "epoch 1632 completed in 27.23 seconds with avg training loss 306.4\n",
      "epoch 1633 completed in 27.06 seconds with avg training loss 306.2\n",
      "epoch 1634 completed in 27.0 seconds with avg training loss 306.56\n",
      "epoch 1635 completed in 26.75 seconds with avg training loss 305.88\n",
      "model not saved. validation loss was  806.1427148437499\n",
      "epoch 1636 completed in 27.03 seconds with avg training loss 305.96\n",
      "epoch 1637 completed in 26.53 seconds with avg training loss 306.35\n",
      "epoch 1638 completed in 27.56 seconds with avg training loss 306.13\n",
      "epoch 1639 completed in 27.23 seconds with avg training loss 306.45\n",
      "epoch 1640 completed in 27.27 seconds with avg training loss 306.18\n",
      "model not saved. validation loss was  800.5554882812501\n",
      "epoch 1641 completed in 27.19 seconds with avg training loss 306.02\n",
      "epoch 1642 completed in 27.72 seconds with avg training loss 305.97\n",
      "epoch 1643 completed in 27.2 seconds with avg training loss 306.29\n",
      "epoch 1644 completed in 27.72 seconds with avg training loss 305.88\n",
      "epoch 1645 completed in 27.31 seconds with avg training loss 306.59\n",
      "model not saved. validation loss was  804.11671875\n",
      "epoch 1646 completed in 27.25 seconds with avg training loss 306.65\n",
      "epoch 1647 completed in 27.64 seconds with avg training loss 305.65\n",
      "epoch 1648 completed in 27.23 seconds with avg training loss 306.85\n",
      "epoch 1649 completed in 27.09 seconds with avg training loss 306.41\n",
      "epoch 1650 completed in 27.06 seconds with avg training loss 306.28\n",
      "model not saved. validation loss was  800.97888671875\n",
      "epoch 1651 completed in 27.36 seconds with avg training loss 305.95\n",
      "epoch 1652 completed in 26.94 seconds with avg training loss 305.96\n",
      "epoch 1653 completed in 27.42 seconds with avg training loss 305.95\n",
      "epoch 1654 completed in 27.33 seconds with avg training loss 306.08\n",
      "epoch 1655 completed in 27.17 seconds with avg training loss 306.36\n",
      "model not saved. validation loss was  800.94650390625\n",
      "epoch 1656 completed in 27.11 seconds with avg training loss 306.28\n",
      "epoch 1657 completed in 27.23 seconds with avg training loss 306.0\n",
      "epoch 1658 completed in 27.14 seconds with avg training loss 305.4\n",
      "epoch 1659 completed in 27.45 seconds with avg training loss 306.03\n",
      "epoch 1660 completed in 26.89 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  803.6331249999998\n",
      "epoch 1661 completed in 27.44 seconds with avg training loss 306.46\n",
      "epoch 1662 completed in 27.5 seconds with avg training loss 306.39\n",
      "epoch 1663 completed in 27.11 seconds with avg training loss 306.39\n",
      "epoch 1664 completed in 27.56 seconds with avg training loss 306.38\n",
      "epoch 1665 completed in 27.66 seconds with avg training loss 306.06\n",
      "model not saved. validation loss was  808.7598046875\n",
      "epoch 1666 completed in 27.33 seconds with avg training loss 306.64\n",
      "epoch 1667 completed in 26.89 seconds with avg training loss 306.32\n",
      "epoch 1668 completed in 26.86 seconds with avg training loss 306.22\n",
      "epoch 1669 completed in 26.81 seconds with avg training loss 306.58\n",
      "epoch 1670 completed in 27.08 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  804.67724609375\n",
      "epoch 1671 completed in 27.23 seconds with avg training loss 306.04\n",
      "epoch 1672 completed in 27.14 seconds with avg training loss 305.9\n",
      "epoch 1673 completed in 27.17 seconds with avg training loss 306.34\n",
      "epoch 1674 completed in 27.47 seconds with avg training loss 306.19\n",
      "epoch 1675 completed in 27.23 seconds with avg training loss 305.83\n",
      "model not saved. validation loss was  805.2669726562501\n",
      "epoch 1676 completed in 27.42 seconds with avg training loss 306.63\n",
      "epoch 1677 completed in 27.25 seconds with avg training loss 305.93\n",
      "epoch 1678 completed in 27.39 seconds with avg training loss 306.13\n",
      "epoch 1679 completed in 27.55 seconds with avg training loss 306.44\n",
      "epoch 1680 completed in 27.23 seconds with avg training loss 306.53\n",
      "model not saved. validation loss was  802.734609375\n",
      "epoch 1681 completed in 27.14 seconds with avg training loss 306.47\n",
      "epoch 1682 completed in 26.98 seconds with avg training loss 306.32\n",
      "epoch 1683 completed in 27.39 seconds with avg training loss 306.17\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1684 completed in 27.14 seconds with avg training loss 306.17\n",
      "epoch 1685 completed in 26.97 seconds with avg training loss 306.09\n",
      "model not saved. validation loss was  813.27958984375\n",
      "epoch 1686 completed in 27.33 seconds with avg training loss 305.95\n",
      "epoch 1687 completed in 27.36 seconds with avg training loss 305.94\n",
      "epoch 1688 completed in 27.36 seconds with avg training loss 305.83\n",
      "epoch 1689 completed in 27.34 seconds with avg training loss 306.61\n",
      "epoch 1690 completed in 28.09 seconds with avg training loss 306.51\n",
      "model not saved. validation loss was  806.016953125\n",
      "epoch 1691 completed in 27.48 seconds with avg training loss 306.16\n",
      "epoch 1692 completed in 27.08 seconds with avg training loss 306.09\n",
      "epoch 1693 completed in 27.2 seconds with avg training loss 306.43\n",
      "epoch 1694 completed in 27.16 seconds with avg training loss 306.2\n",
      "epoch 1695 completed in 27.06 seconds with avg training loss 306.23\n",
      "model not saved. validation loss was  798.96240234375\n",
      "epoch 1696 completed in 27.19 seconds with avg training loss 306.26\n",
      "epoch 1697 completed in 26.56 seconds with avg training loss 306.13\n",
      "epoch 1698 completed in 26.66 seconds with avg training loss 306.85\n",
      "epoch 1699 completed in 27.27 seconds with avg training loss 306.19\n",
      "epoch 1700 completed in 27.08 seconds with avg training loss 306.05\n",
      "model not saved. validation loss was  805.69533203125\n",
      "epoch 1701 completed in 26.61 seconds with avg training loss 305.87\n",
      "epoch 1702 completed in 26.67 seconds with avg training loss 306.06\n",
      "epoch 1703 completed in 26.8 seconds with avg training loss 306.41\n",
      "epoch 1704 completed in 27.2 seconds with avg training loss 306.21\n",
      "epoch 1705 completed in 26.84 seconds with avg training loss 306.37\n",
      "model not saved. validation loss was  808.815\n",
      "epoch 1706 completed in 26.47 seconds with avg training loss 305.59\n",
      "epoch 1707 completed in 27.09 seconds with avg training loss 306.09\n",
      "epoch 1708 completed in 26.73 seconds with avg training loss 305.78\n",
      "epoch 1709 completed in 26.77 seconds with avg training loss 306.23\n",
      "epoch 1710 completed in 26.61 seconds with avg training loss 305.27\n",
      "model not saved. validation loss was  802.56318359375\n",
      "epoch 1711 completed in 27.08 seconds with avg training loss 306.6\n",
      "epoch 1712 completed in 27.06 seconds with avg training loss 305.94\n",
      "epoch 1713 completed in 26.88 seconds with avg training loss 306.09\n",
      "epoch 1714 completed in 27.06 seconds with avg training loss 306.57\n",
      "epoch 1715 completed in 26.47 seconds with avg training loss 306.37\n",
      "model not saved. validation loss was  814.338671875\n",
      "epoch 1716 completed in 27.08 seconds with avg training loss 306.59\n",
      "epoch 1717 completed in 26.84 seconds with avg training loss 306.45\n",
      "epoch 1718 completed in 27.16 seconds with avg training loss 306.27\n",
      "epoch 1719 completed in 27.45 seconds with avg training loss 305.89\n",
      "epoch 1720 completed in 27.27 seconds with avg training loss 305.93\n",
      "model not saved. validation loss was  803.03900390625\n",
      "epoch 1721 completed in 27.23 seconds with avg training loss 306.66\n",
      "epoch 1722 completed in 27.17 seconds with avg training loss 306.18\n",
      "epoch 1723 completed in 27.94 seconds with avg training loss 305.75\n",
      "epoch 1724 completed in 27.42 seconds with avg training loss 306.21\n",
      "epoch 1725 completed in 26.91 seconds with avg training loss 306.32\n",
      "model not saved. validation loss was  807.6440625\n",
      "epoch 1726 completed in 27.64 seconds with avg training loss 305.9\n",
      "epoch 1727 completed in 27.03 seconds with avg training loss 305.96\n",
      "epoch 1728 completed in 27.08 seconds with avg training loss 306.21\n",
      "epoch 1729 completed in 26.92 seconds with avg training loss 306.22\n",
      "epoch 1730 completed in 26.64 seconds with avg training loss 306.21\n",
      "model not saved. validation loss was  799.5606250000001\n",
      "epoch 1731 completed in 26.77 seconds with avg training loss 306.37\n",
      "epoch 1732 completed in 26.55 seconds with avg training loss 306.23\n",
      "epoch 1733 completed in 26.7 seconds with avg training loss 306.74\n",
      "epoch 1734 completed in 27.06 seconds with avg training loss 306.49\n",
      "epoch 1735 completed in 27.39 seconds with avg training loss 305.56\n",
      "model not saved. validation loss was  804.78736328125\n",
      "epoch 1736 completed in 27.55 seconds with avg training loss 305.77\n",
      "epoch 1737 completed in 27.19 seconds with avg training loss 305.58\n",
      "epoch 1738 completed in 26.88 seconds with avg training loss 306.21\n",
      "epoch 1739 completed in 26.66 seconds with avg training loss 305.87\n",
      "epoch 1740 completed in 26.86 seconds with avg training loss 306.18\n",
      "model not saved. validation loss was  809.5118164062501\n",
      "epoch 1741 completed in 26.95 seconds with avg training loss 306.36\n",
      "epoch 1742 completed in 26.89 seconds with avg training loss 305.97\n",
      "epoch 1743 completed in 27.11 seconds with avg training loss 306.27\n",
      "epoch 1744 completed in 26.84 seconds with avg training loss 306.32\n",
      "epoch 1745 completed in 26.86 seconds with avg training loss 305.71\n",
      "model not saved. validation loss was  810.0621874999999\n",
      "epoch 1746 completed in 27.17 seconds with avg training loss 306.5\n",
      "epoch 1747 completed in 27.19 seconds with avg training loss 306.9\n",
      "epoch 1748 completed in 27.0 seconds with avg training loss 306.29\n",
      "epoch 1749 completed in 27.25 seconds with avg training loss 306.2\n",
      "epoch 1750 completed in 27.17 seconds with avg training loss 306.42\n",
      "model not saved. validation loss was  808.1051562499999\n",
      "epoch 1751 completed in 27.09 seconds with avg training loss 306.02\n",
      "epoch 1752 completed in 27.44 seconds with avg training loss 305.82\n",
      "epoch 1753 completed in 26.44 seconds with avg training loss 306.6\n",
      "epoch 1754 completed in 27.31 seconds with avg training loss 306.93\n",
      "epoch 1755 completed in 27.13 seconds with avg training loss 306.53\n",
      "model not saved. validation loss was  805.94869140625\n",
      "epoch 1756 completed in 26.67 seconds with avg training loss 306.38\n",
      "epoch 1757 completed in 26.81 seconds with avg training loss 306.62\n",
      "epoch 1758 completed in 27.03 seconds with avg training loss 306.13\n",
      "epoch 1759 completed in 27.11 seconds with avg training loss 306.2\n",
      "epoch 1760 completed in 27.0 seconds with avg training loss 306.31\n",
      "model not saved. validation loss was  807.82392578125\n",
      "epoch 1761 completed in 26.7 seconds with avg training loss 305.97\n",
      "epoch 1762 completed in 26.56 seconds with avg training loss 306.37\n",
      "epoch 1763 completed in 27.22 seconds with avg training loss 306.85\n",
      "epoch 1764 completed in 27.17 seconds with avg training loss 305.77\n",
      "epoch 1765 completed in 26.8 seconds with avg training loss 305.85\n",
      "model not saved. validation loss was  806.7095507812501\n",
      "epoch 1766 completed in 27.09 seconds with avg training loss 306.13\n",
      "epoch 1767 completed in 26.94 seconds with avg training loss 306.25\n",
      "epoch 1768 completed in 27.03 seconds with avg training loss 306.29\n",
      "epoch 1769 completed in 26.97 seconds with avg training loss 306.1\n",
      "epoch 1770 completed in 26.78 seconds with avg training loss 306.82\n",
      "model not saved. validation loss was  800.5877929687499\n",
      "epoch 1771 completed in 26.66 seconds with avg training loss 306.73\n",
      "epoch 1772 completed in 27.2 seconds with avg training loss 305.87\n",
      "epoch 1773 completed in 26.45 seconds with avg training loss 305.94\n",
      "epoch 1774 completed in 26.78 seconds with avg training loss 305.82\n",
      "epoch 1775 completed in 26.73 seconds with avg training loss 306.36\n",
      "model not saved. validation loss was  807.9896875\n",
      "epoch 1776 completed in 27.14 seconds with avg training loss 306.11\n",
      "epoch 1777 completed in 27.11 seconds with avg training loss 306.07\n",
      "epoch 1778 completed in 26.8 seconds with avg training loss 305.85\n",
      "epoch 1779 completed in 27.02 seconds with avg training loss 305.79\n",
      "epoch 1780 completed in 27.47 seconds with avg training loss 306.77\n",
      "model not saved. validation loss was  809.89849609375\n",
      "epoch 1781 completed in 27.5 seconds with avg training loss 306.31\n",
      "epoch 1782 completed in 27.69 seconds with avg training loss 305.83\n",
      "epoch 1783 completed in 27.13 seconds with avg training loss 306.37\n",
      "epoch 1784 completed in 27.19 seconds with avg training loss 306.45\n",
      "epoch 1785 completed in 27.2 seconds with avg training loss 306.55\n",
      "model not saved. validation loss was  805.96724609375\n",
      "epoch 1786 completed in 27.0 seconds with avg training loss 306.1\n",
      "epoch 1787 completed in 26.98 seconds with avg training loss 306.16\n",
      "epoch 1788 completed in 27.48 seconds with avg training loss 306.27\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1789 completed in 27.0 seconds with avg training loss 306.49\n",
      "epoch 1790 completed in 27.19 seconds with avg training loss 306.43\n",
      "model not saved. validation loss was  812.7578515625\n",
      "epoch 1791 completed in 26.89 seconds with avg training loss 306.1\n",
      "epoch 1792 completed in 26.83 seconds with avg training loss 306.0\n",
      "epoch 1793 completed in 26.86 seconds with avg training loss 306.16\n",
      "epoch 1794 completed in 26.86 seconds with avg training loss 305.72\n",
      "epoch 1795 completed in 27.2 seconds with avg training loss 306.31\n",
      "model not saved. validation loss was  805.85900390625\n",
      "epoch 1796 completed in 27.31 seconds with avg training loss 306.15\n",
      "epoch 1797 completed in 27.2 seconds with avg training loss 306.27\n",
      "epoch 1798 completed in 27.09 seconds with avg training loss 306.58\n",
      "epoch 1799 completed in 27.48 seconds with avg training loss 306.08\n",
      "epoch 1800 completed in 26.94 seconds with avg training loss 306.33\n",
      "model not saved. validation loss was  801.75322265625\n",
      "epoch 1801 completed in 27.17 seconds with avg training loss 305.86\n",
      "epoch 1802 completed in 27.17 seconds with avg training loss 306.6\n",
      "epoch 1803 completed in 26.75 seconds with avg training loss 306.3\n",
      "epoch 1804 completed in 27.47 seconds with avg training loss 306.32\n",
      "epoch 1805 completed in 27.64 seconds with avg training loss 306.23\n",
      "model not saved. validation loss was  798.203671875\n",
      "epoch 1806 completed in 27.52 seconds with avg training loss 306.27\n",
      "epoch 1807 completed in 26.8 seconds with avg training loss 306.18\n",
      "epoch 1808 completed in 26.75 seconds with avg training loss 306.1\n",
      "epoch 1809 completed in 27.19 seconds with avg training loss 306.17\n",
      "epoch 1810 completed in 27.06 seconds with avg training loss 306.09\n",
      "model not saved. validation loss was  805.4688867187501\n",
      "epoch 1811 completed in 26.53 seconds with avg training loss 306.58\n",
      "epoch 1812 completed in 26.61 seconds with avg training loss 306.22\n",
      "epoch 1813 completed in 27.06 seconds with avg training loss 306.01\n",
      "epoch 1814 completed in 27.33 seconds with avg training loss 306.37\n",
      "epoch 1815 completed in 27.16 seconds with avg training loss 306.25\n",
      "model not saved. validation loss was  802.7611132812499\n",
      "epoch 1816 completed in 26.81 seconds with avg training loss 306.47\n",
      "epoch 1817 completed in 26.67 seconds with avg training loss 306.31\n",
      "epoch 1818 completed in 27.0 seconds with avg training loss 306.81\n",
      "epoch 1819 completed in 26.58 seconds with avg training loss 307.01\n",
      "epoch 1820 completed in 27.3 seconds with avg training loss 306.11\n",
      "model not saved. validation loss was  807.522265625\n",
      "epoch 1821 completed in 27.17 seconds with avg training loss 306.5\n",
      "epoch 1822 completed in 27.36 seconds with avg training loss 306.09\n",
      "epoch 1823 completed in 27.28 seconds with avg training loss 306.54\n",
      "epoch 1824 completed in 27.3 seconds with avg training loss 305.47\n",
      "epoch 1825 completed in 27.0 seconds with avg training loss 306.02\n",
      "model not saved. validation loss was  801.33326171875\n",
      "epoch 1826 completed in 27.19 seconds with avg training loss 306.15\n",
      "epoch 1827 completed in 27.17 seconds with avg training loss 305.63\n",
      "epoch 1828 completed in 26.83 seconds with avg training loss 306.64\n",
      "epoch 1829 completed in 26.78 seconds with avg training loss 305.7\n",
      "epoch 1830 completed in 26.45 seconds with avg training loss 306.31\n",
      "model not saved. validation loss was  807.95771484375\n",
      "epoch 1831 completed in 26.45 seconds with avg training loss 305.91\n",
      "epoch 1832 completed in 26.94 seconds with avg training loss 306.28\n",
      "epoch 1833 completed in 26.31 seconds with avg training loss 306.25\n",
      "epoch 1834 completed in 26.44 seconds with avg training loss 306.44\n",
      "epoch 1835 completed in 27.48 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  809.7530859374999\n",
      "epoch 1836 completed in 27.16 seconds with avg training loss 306.36\n",
      "epoch 1837 completed in 27.14 seconds with avg training loss 305.91\n",
      "epoch 1838 completed in 27.08 seconds with avg training loss 306.17\n",
      "epoch 1839 completed in 27.73 seconds with avg training loss 306.2\n",
      "epoch 1840 completed in 27.38 seconds with avg training loss 306.82\n",
      "model not saved. validation loss was  806.78326171875\n",
      "epoch 1841 completed in 26.92 seconds with avg training loss 306.01\n",
      "epoch 1842 completed in 26.94 seconds with avg training loss 306.32\n",
      "epoch 1843 completed in 26.92 seconds with avg training loss 306.21\n",
      "epoch 1844 completed in 26.67 seconds with avg training loss 306.4\n",
      "epoch 1845 completed in 26.95 seconds with avg training loss 305.88\n",
      "model not saved. validation loss was  811.15013671875\n",
      "epoch 1846 completed in 26.52 seconds with avg training loss 306.35\n",
      "epoch 1847 completed in 26.95 seconds with avg training loss 306.59\n",
      "epoch 1848 completed in 26.59 seconds with avg training loss 306.79\n",
      "epoch 1849 completed in 26.63 seconds with avg training loss 306.09\n",
      "epoch 1850 completed in 26.61 seconds with avg training loss 306.28\n",
      "model not saved. validation loss was  802.826796875\n",
      "epoch 1851 completed in 27.03 seconds with avg training loss 305.77\n",
      "epoch 1852 completed in 27.28 seconds with avg training loss 306.43\n",
      "epoch 1853 completed in 26.94 seconds with avg training loss 306.12\n",
      "epoch 1854 completed in 26.94 seconds with avg training loss 306.34\n",
      "epoch 1855 completed in 27.64 seconds with avg training loss 306.46\n",
      "model not saved. validation loss was  812.8643359375001\n",
      "epoch 1856 completed in 27.41 seconds with avg training loss 306.54\n",
      "epoch 1857 completed in 26.83 seconds with avg training loss 305.77\n",
      "epoch 1858 completed in 27.22 seconds with avg training loss 306.28\n",
      "epoch 1859 completed in 26.52 seconds with avg training loss 306.06\n",
      "epoch 1860 completed in 26.98 seconds with avg training loss 306.18\n",
      "model not saved. validation loss was  804.9202734375001\n",
      "epoch 1861 completed in 26.88 seconds with avg training loss 306.24\n",
      "epoch 1862 completed in 26.66 seconds with avg training loss 306.61\n",
      "epoch 1863 completed in 26.78 seconds with avg training loss 305.55\n",
      "epoch 1864 completed in 26.75 seconds with avg training loss 305.94\n",
      "epoch 1865 completed in 27.17 seconds with avg training loss 305.6\n",
      "model not saved. validation loss was  804.82193359375\n",
      "epoch 1866 completed in 26.56 seconds with avg training loss 306.22\n",
      "epoch 1867 completed in 27.03 seconds with avg training loss 305.57\n",
      "epoch 1868 completed in 26.91 seconds with avg training loss 305.92\n",
      "epoch 1869 completed in 26.5 seconds with avg training loss 306.18\n",
      "epoch 1870 completed in 26.48 seconds with avg training loss 306.39\n",
      "model not saved. validation loss was  804.47416015625\n",
      "epoch 1871 completed in 26.94 seconds with avg training loss 306.3\n",
      "epoch 1872 completed in 26.73 seconds with avg training loss 306.16\n",
      "epoch 1873 completed in 26.98 seconds with avg training loss 306.32\n",
      "epoch 1874 completed in 27.44 seconds with avg training loss 306.06\n",
      "epoch 1875 completed in 26.81 seconds with avg training loss 306.11\n",
      "model not saved. validation loss was  803.2559375000001\n",
      "epoch 1876 completed in 27.64 seconds with avg training loss 305.77\n",
      "epoch 1877 completed in 27.52 seconds with avg training loss 306.42\n",
      "epoch 1878 completed in 27.38 seconds with avg training loss 306.53\n",
      "epoch 1879 completed in 27.05 seconds with avg training loss 306.05\n",
      "epoch 1880 completed in 26.89 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  811.8291796875001\n",
      "epoch 1881 completed in 26.91 seconds with avg training loss 306.64\n",
      "epoch 1882 completed in 27.22 seconds with avg training loss 305.94\n",
      "epoch 1883 completed in 27.13 seconds with avg training loss 306.2\n",
      "epoch 1884 completed in 26.61 seconds with avg training loss 306.19\n",
      "epoch 1885 completed in 27.16 seconds with avg training loss 306.77\n",
      "model not saved. validation loss was  808.81912109375\n",
      "epoch 1886 completed in 26.94 seconds with avg training loss 305.96\n",
      "epoch 1887 completed in 26.63 seconds with avg training loss 306.26\n",
      "epoch 1888 completed in 26.63 seconds with avg training loss 306.37\n",
      "epoch 1889 completed in 26.91 seconds with avg training loss 306.35\n",
      "epoch 1890 completed in 26.81 seconds with avg training loss 305.85\n",
      "model not saved. validation loss was  798.3821484375\n",
      "epoch 1891 completed in 26.84 seconds with avg training loss 306.64\n",
      "epoch 1892 completed in 27.14 seconds with avg training loss 305.91\n",
      "epoch 1893 completed in 27.33 seconds with avg training loss 306.14\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1894 completed in 27.33 seconds with avg training loss 306.26\n",
      "epoch 1895 completed in 26.66 seconds with avg training loss 306.06\n",
      "model not saved. validation loss was  810.2762695312499\n",
      "epoch 1896 completed in 27.03 seconds with avg training loss 306.39\n",
      "epoch 1897 completed in 26.77 seconds with avg training loss 306.39\n",
      "epoch 1898 completed in 26.45 seconds with avg training loss 306.3\n",
      "epoch 1899 completed in 26.67 seconds with avg training loss 306.43\n",
      "epoch 1900 completed in 26.88 seconds with avg training loss 306.24\n",
      "model not saved. validation loss was  802.8199999999999\n",
      "epoch 1901 completed in 26.44 seconds with avg training loss 306.55\n",
      "epoch 1902 completed in 26.42 seconds with avg training loss 306.38\n",
      "epoch 1903 completed in 26.61 seconds with avg training loss 305.94\n",
      "epoch 1904 completed in 26.55 seconds with avg training loss 306.87\n",
      "epoch 1905 completed in 27.05 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  806.1582421875\n",
      "epoch 1906 completed in 27.02 seconds with avg training loss 306.24\n",
      "epoch 1907 completed in 27.34 seconds with avg training loss 306.3\n",
      "epoch 1908 completed in 27.0 seconds with avg training loss 306.19\n",
      "epoch 1909 completed in 27.64 seconds with avg training loss 306.22\n",
      "epoch 1910 completed in 27.22 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  808.8658789062499\n",
      "epoch 1911 completed in 27.44 seconds with avg training loss 306.31\n",
      "epoch 1912 completed in 27.16 seconds with avg training loss 305.65\n",
      "epoch 1913 completed in 27.0 seconds with avg training loss 306.16\n",
      "epoch 1914 completed in 26.94 seconds with avg training loss 306.01\n",
      "epoch 1915 completed in 27.25 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  803.6260351562499\n",
      "epoch 1916 completed in 27.3 seconds with avg training loss 306.16\n",
      "epoch 1917 completed in 27.03 seconds with avg training loss 305.84\n",
      "epoch 1918 completed in 26.73 seconds with avg training loss 305.86\n",
      "epoch 1919 completed in 27.23 seconds with avg training loss 305.64\n",
      "epoch 1920 completed in 27.45 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  808.3937500000001\n",
      "epoch 1921 completed in 27.27 seconds with avg training loss 306.81\n",
      "epoch 1922 completed in 27.03 seconds with avg training loss 306.43\n",
      "epoch 1923 completed in 27.34 seconds with avg training loss 306.14\n",
      "epoch 1924 completed in 27.09 seconds with avg training loss 306.07\n",
      "epoch 1925 completed in 27.06 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  806.08970703125\n",
      "epoch 1926 completed in 27.02 seconds with avg training loss 306.22\n",
      "epoch 1927 completed in 26.7 seconds with avg training loss 305.92\n",
      "epoch 1928 completed in 26.7 seconds with avg training loss 305.96\n",
      "epoch 1929 completed in 26.45 seconds with avg training loss 306.5\n",
      "epoch 1930 completed in 26.11 seconds with avg training loss 305.92\n",
      "model not saved. validation loss was  809.60716796875\n",
      "epoch 1931 completed in 26.5 seconds with avg training loss 305.92\n",
      "epoch 1932 completed in 27.13 seconds with avg training loss 306.04\n",
      "epoch 1933 completed in 27.41 seconds with avg training loss 306.19\n",
      "epoch 1934 completed in 27.39 seconds with avg training loss 306.42\n",
      "epoch 1935 completed in 26.95 seconds with avg training loss 306.2\n",
      "model not saved. validation loss was  808.5392578125\n",
      "epoch 1936 completed in 27.28 seconds with avg training loss 306.27\n",
      "epoch 1937 completed in 26.86 seconds with avg training loss 305.85\n",
      "epoch 1938 completed in 26.34 seconds with avg training loss 306.15\n",
      "epoch 1939 completed in 26.83 seconds with avg training loss 306.19\n",
      "epoch 1940 completed in 26.84 seconds with avg training loss 305.94\n",
      "model not saved. validation loss was  806.953046875\n",
      "epoch 1941 completed in 26.8 seconds with avg training loss 306.34\n",
      "epoch 1942 completed in 27.48 seconds with avg training loss 305.9\n",
      "epoch 1943 completed in 26.95 seconds with avg training loss 306.32\n",
      "epoch 1944 completed in 27.23 seconds with avg training loss 306.01\n",
      "epoch 1945 completed in 26.8 seconds with avg training loss 306.56\n",
      "model not saved. validation loss was  804.205859375\n",
      "epoch 1946 completed in 26.89 seconds with avg training loss 306.44\n",
      "epoch 1947 completed in 27.23 seconds with avg training loss 305.51\n",
      "epoch 1948 completed in 27.02 seconds with avg training loss 306.23\n",
      "epoch 1949 completed in 27.42 seconds with avg training loss 306.11\n",
      "epoch 1950 completed in 26.58 seconds with avg training loss 306.27\n",
      "model not saved. validation loss was  805.8225390625\n",
      "epoch 1951 completed in 26.73 seconds with avg training loss 306.32\n",
      "epoch 1952 completed in 26.94 seconds with avg training loss 306.6\n",
      "epoch 1953 completed in 26.95 seconds with avg training loss 306.3\n",
      "epoch 1954 completed in 26.77 seconds with avg training loss 305.78\n",
      "epoch 1955 completed in 26.33 seconds with avg training loss 306.35\n",
      "model not saved. validation loss was  807.0945898437501\n",
      "epoch 1956 completed in 26.59 seconds with avg training loss 306.19\n",
      "epoch 1957 completed in 26.88 seconds with avg training loss 305.6\n",
      "epoch 1958 completed in 27.03 seconds with avg training loss 306.37\n",
      "epoch 1959 completed in 27.36 seconds with avg training loss 306.07\n",
      "epoch 1960 completed in 27.08 seconds with avg training loss 306.15\n",
      "model not saved. validation loss was  809.5762890625\n",
      "epoch 1961 completed in 26.53 seconds with avg training loss 306.44\n",
      "epoch 1962 completed in 26.47 seconds with avg training loss 306.26\n",
      "epoch 1963 completed in 26.64 seconds with avg training loss 305.95\n",
      "epoch 1964 completed in 27.19 seconds with avg training loss 306.74\n",
      "epoch 1965 completed in 27.17 seconds with avg training loss 306.54\n",
      "model not saved. validation loss was  808.2125585937499\n",
      "epoch 1966 completed in 26.89 seconds with avg training loss 306.22\n",
      "epoch 1967 completed in 26.53 seconds with avg training loss 306.32\n",
      "epoch 1968 completed in 26.3 seconds with avg training loss 306.09\n",
      "epoch 1969 completed in 26.61 seconds with avg training loss 306.31\n",
      "epoch 1970 completed in 26.5 seconds with avg training loss 306.24\n",
      "model not saved. validation loss was  806.11658203125\n",
      "epoch 1971 completed in 26.58 seconds with avg training loss 305.61\n",
      "epoch 1972 completed in 26.67 seconds with avg training loss 305.95\n",
      "epoch 1973 completed in 26.89 seconds with avg training loss 305.88\n",
      "epoch 1974 completed in 26.91 seconds with avg training loss 306.18\n",
      "epoch 1975 completed in 26.8 seconds with avg training loss 305.91\n",
      "model not saved. validation loss was  802.1234375\n",
      "epoch 1976 completed in 26.5 seconds with avg training loss 305.85\n",
      "epoch 1977 completed in 27.31 seconds with avg training loss 306.16\n",
      "epoch 1978 completed in 27.2 seconds with avg training loss 306.39\n",
      "epoch 1979 completed in 26.89 seconds with avg training loss 306.07\n",
      "epoch 1980 completed in 27.3 seconds with avg training loss 305.81\n",
      "model not saved. validation loss was  807.63380859375\n",
      "epoch 1981 completed in 26.73 seconds with avg training loss 306.03\n",
      "epoch 1982 completed in 26.77 seconds with avg training loss 306.4\n",
      "epoch 1983 completed in 27.25 seconds with avg training loss 305.94\n",
      "epoch 1984 completed in 26.92 seconds with avg training loss 305.86\n",
      "epoch 1985 completed in 27.11 seconds with avg training loss 305.83\n",
      "model not saved. validation loss was  800.9576953125\n",
      "epoch 1986 completed in 27.06 seconds with avg training loss 306.66\n",
      "epoch 1987 completed in 27.16 seconds with avg training loss 306.05\n",
      "epoch 1988 completed in 27.11 seconds with avg training loss 306.09\n",
      "epoch 1989 completed in 27.11 seconds with avg training loss 305.94\n",
      "epoch 1990 completed in 27.27 seconds with avg training loss 305.66\n",
      "model not saved. validation loss was  803.84580078125\n",
      "epoch 1991 completed in 27.3 seconds with avg training loss 306.51\n",
      "epoch 1992 completed in 27.27 seconds with avg training loss 305.81\n",
      "epoch 1993 completed in 26.94 seconds with avg training loss 305.86\n",
      "epoch 1994 completed in 26.97 seconds with avg training loss 306.21\n",
      "epoch 1995 completed in 26.92 seconds with avg training loss 306.27\n",
      "model not saved. validation loss was  803.8815039062499\n",
      "epoch 1996 completed in 27.28 seconds with avg training loss 306.58\n",
      "epoch 1997 completed in 27.05 seconds with avg training loss 305.86\n",
      "epoch 1998 completed in 27.28 seconds with avg training loss 305.85\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1999 completed in 27.39 seconds with avg training loss 306.46\n",
      "epoch 2000 completed in 27.11 seconds with avg training loss 306.18\n",
      "model not saved. validation loss was  805.96783203125\n",
      "epoch 2001 completed in 26.94 seconds with avg training loss 305.94\n",
      "epoch 2002 completed in 27.45 seconds with avg training loss 305.83\n",
      "epoch 2003 completed in 26.75 seconds with avg training loss 306.6\n",
      "epoch 2004 completed in 26.44 seconds with avg training loss 306.41\n",
      "epoch 2005 completed in 26.44 seconds with avg training loss 306.32\n",
      "model not saved. validation loss was  814.9102734374999\n",
      "epoch 2006 completed in 26.75 seconds with avg training loss 306.07\n",
      "epoch 2007 completed in 26.84 seconds with avg training loss 306.13\n",
      "epoch 2008 completed in 26.86 seconds with avg training loss 306.26\n",
      "epoch 2009 completed in 27.14 seconds with avg training loss 306.11\n",
      "epoch 2010 completed in 26.75 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  808.1033789062501\n",
      "epoch 2011 completed in 26.73 seconds with avg training loss 306.22\n",
      "epoch 2012 completed in 27.03 seconds with avg training loss 305.79\n",
      "epoch 2013 completed in 26.91 seconds with avg training loss 306.64\n",
      "epoch 2014 completed in 26.98 seconds with avg training loss 307.11\n",
      "epoch 2015 completed in 27.0 seconds with avg training loss 306.39\n",
      "model not saved. validation loss was  805.12716796875\n",
      "epoch 2016 completed in 27.27 seconds with avg training loss 305.99\n",
      "epoch 2017 completed in 26.64 seconds with avg training loss 306.25\n",
      "epoch 2018 completed in 26.91 seconds with avg training loss 306.5\n",
      "epoch 2019 completed in 26.67 seconds with avg training loss 306.53\n",
      "epoch 2020 completed in 27.48 seconds with avg training loss 305.79\n",
      "model not saved. validation loss was  805.112734375\n",
      "epoch 2021 completed in 26.8 seconds with avg training loss 306.02\n",
      "epoch 2022 completed in 26.61 seconds with avg training loss 305.82\n",
      "epoch 2023 completed in 26.66 seconds with avg training loss 306.33\n",
      "epoch 2024 completed in 26.48 seconds with avg training loss 306.21\n",
      "epoch 2025 completed in 26.25 seconds with avg training loss 306.09\n",
      "model not saved. validation loss was  808.5475390624999\n",
      "epoch 2026 completed in 26.92 seconds with avg training loss 306.09\n",
      "epoch 2027 completed in 26.89 seconds with avg training loss 306.45\n",
      "epoch 2028 completed in 27.23 seconds with avg training loss 306.16\n",
      "epoch 2029 completed in 27.0 seconds with avg training loss 306.02\n",
      "epoch 2030 completed in 26.59 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  808.17068359375\n",
      "epoch 2031 completed in 26.78 seconds with avg training loss 306.06\n",
      "epoch 2032 completed in 27.0 seconds with avg training loss 306.44\n",
      "epoch 2033 completed in 27.02 seconds with avg training loss 306.41\n",
      "epoch 2034 completed in 26.78 seconds with avg training loss 306.43\n",
      "epoch 2035 completed in 26.81 seconds with avg training loss 306.29\n",
      "model not saved. validation loss was  805.8871484375\n",
      "epoch 2036 completed in 26.61 seconds with avg training loss 306.04\n",
      "epoch 2037 completed in 26.92 seconds with avg training loss 306.15\n",
      "epoch 2038 completed in 26.88 seconds with avg training loss 306.06\n",
      "epoch 2039 completed in 27.25 seconds with avg training loss 306.35\n",
      "epoch 2040 completed in 27.39 seconds with avg training loss 306.29\n",
      "model not saved. validation loss was  806.6697265624999\n",
      "epoch 2041 completed in 27.03 seconds with avg training loss 306.26\n",
      "epoch 2042 completed in 27.19 seconds with avg training loss 306.17\n",
      "epoch 2043 completed in 27.25 seconds with avg training loss 306.68\n",
      "epoch 2044 completed in 26.78 seconds with avg training loss 306.13\n",
      "epoch 2045 completed in 26.7 seconds with avg training loss 306.21\n",
      "model not saved. validation loss was  809.07865234375\n",
      "epoch 2046 completed in 26.58 seconds with avg training loss 306.19\n",
      "epoch 2047 completed in 26.75 seconds with avg training loss 306.62\n",
      "epoch 2048 completed in 26.83 seconds with avg training loss 306.13\n",
      "epoch 2049 completed in 27.17 seconds with avg training loss 305.94\n",
      "epoch 2050 completed in 26.89 seconds with avg training loss 306.24\n",
      "model not saved. validation loss was  805.5032421875001\n",
      "epoch 2051 completed in 26.81 seconds with avg training loss 305.8\n",
      "epoch 2052 completed in 26.61 seconds with avg training loss 306.17\n",
      "epoch 2053 completed in 26.52 seconds with avg training loss 306.48\n",
      "epoch 2054 completed in 26.42 seconds with avg training loss 306.19\n",
      "epoch 2055 completed in 26.95 seconds with avg training loss 305.86\n",
      "model not saved. validation loss was  806.9828124999999\n",
      "epoch 2056 completed in 26.88 seconds with avg training loss 306.75\n",
      "epoch 2057 completed in 27.23 seconds with avg training loss 306.1\n",
      "epoch 2058 completed in 27.72 seconds with avg training loss 306.19\n",
      "epoch 2059 completed in 27.11 seconds with avg training loss 306.27\n",
      "epoch 2060 completed in 27.08 seconds with avg training loss 305.73\n",
      "model not saved. validation loss was  811.9305859374999\n",
      "epoch 2061 completed in 26.16 seconds with avg training loss 305.85\n",
      "epoch 2062 completed in 26.33 seconds with avg training loss 305.82\n",
      "epoch 2063 completed in 26.55 seconds with avg training loss 306.23\n",
      "epoch 2064 completed in 26.78 seconds with avg training loss 306.07\n",
      "epoch 2065 completed in 26.83 seconds with avg training loss 305.8\n",
      "model not saved. validation loss was  806.778984375\n",
      "epoch 2066 completed in 26.73 seconds with avg training loss 306.11\n",
      "epoch 2067 completed in 26.78 seconds with avg training loss 306.09\n",
      "epoch 2068 completed in 27.03 seconds with avg training loss 305.94\n",
      "epoch 2069 completed in 27.02 seconds with avg training loss 306.15\n",
      "epoch 2070 completed in 27.06 seconds with avg training loss 305.97\n",
      "model not saved. validation loss was  800.30375\n",
      "epoch 2071 completed in 26.97 seconds with avg training loss 306.47\n",
      "epoch 2072 completed in 27.13 seconds with avg training loss 305.91\n",
      "epoch 2073 completed in 27.36 seconds with avg training loss 306.16\n",
      "epoch 2074 completed in 27.06 seconds with avg training loss 305.89\n",
      "epoch 2075 completed in 27.17 seconds with avg training loss 306.52\n",
      "model not saved. validation loss was  810.09591796875\n",
      "epoch 2076 completed in 26.88 seconds with avg training loss 305.98\n",
      "epoch 2077 completed in 27.3 seconds with avg training loss 306.33\n",
      "epoch 2078 completed in 26.95 seconds with avg training loss 306.2\n",
      "epoch 2079 completed in 26.66 seconds with avg training loss 305.69\n",
      "epoch 2080 completed in 27.34 seconds with avg training loss 305.77\n",
      "model not saved. validation loss was  804.5089062500001\n",
      "epoch 2081 completed in 26.94 seconds with avg training loss 306.17\n",
      "epoch 2082 completed in 26.86 seconds with avg training loss 306.26\n",
      "epoch 2083 completed in 27.3 seconds with avg training loss 306.12\n",
      "epoch 2084 completed in 26.92 seconds with avg training loss 305.88\n",
      "epoch 2085 completed in 26.7 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  808.6728906249999\n",
      "epoch 2086 completed in 26.84 seconds with avg training loss 306.4\n",
      "epoch 2087 completed in 26.61 seconds with avg training loss 306.27\n",
      "epoch 2088 completed in 26.69 seconds with avg training loss 306.02\n",
      "epoch 2089 completed in 26.69 seconds with avg training loss 306.36\n",
      "epoch 2090 completed in 27.13 seconds with avg training loss 305.98\n",
      "model not saved. validation loss was  803.3181640625\n",
      "epoch 2091 completed in 26.77 seconds with avg training loss 306.44\n",
      "epoch 2092 completed in 26.78 seconds with avg training loss 306.19\n",
      "epoch 2093 completed in 26.66 seconds with avg training loss 306.11\n",
      "epoch 2094 completed in 26.72 seconds with avg training loss 306.36\n",
      "epoch 2095 completed in 27.02 seconds with avg training loss 305.88\n",
      "model not saved. validation loss was  812.14349609375\n",
      "epoch 2096 completed in 26.75 seconds with avg training loss 306.28\n",
      "epoch 2097 completed in 26.78 seconds with avg training loss 306.58\n",
      "epoch 2098 completed in 27.25 seconds with avg training loss 306.46\n",
      "epoch 2099 completed in 27.31 seconds with avg training loss 306.55\n",
      "epoch 2100 completed in 26.95 seconds with avg training loss 306.41\n",
      "model not saved. validation loss was  798.79775390625\n",
      "epoch 2101 completed in 26.73 seconds with avg training loss 305.58\n",
      "epoch 2102 completed in 26.17 seconds with avg training loss 306.33\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 2103 completed in 26.73 seconds with avg training loss 306.33\n",
      "epoch 2104 completed in 26.69 seconds with avg training loss 306.23\n",
      "epoch 2105 completed in 26.78 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  808.1151757812501\n",
      "epoch 2106 completed in 27.06 seconds with avg training loss 305.52\n",
      "epoch 2107 completed in 27.39 seconds with avg training loss 306.47\n",
      "epoch 2108 completed in 27.09 seconds with avg training loss 306.06\n",
      "epoch 2109 completed in 27.11 seconds with avg training loss 306.2\n",
      "epoch 2110 completed in 27.02 seconds with avg training loss 306.45\n",
      "model not saved. validation loss was  804.8870507812501\n",
      "epoch 2111 completed in 27.25 seconds with avg training loss 306.28\n",
      "epoch 2112 completed in 26.67 seconds with avg training loss 306.17\n",
      "epoch 2113 completed in 27.28 seconds with avg training loss 306.36\n",
      "epoch 2114 completed in 26.8 seconds with avg training loss 306.26\n",
      "epoch 2115 completed in 27.0 seconds with avg training loss 306.39\n",
      "model not saved. validation loss was  794.0578906249999\n",
      "epoch 2116 completed in 26.42 seconds with avg training loss 306.26\n",
      "epoch 2117 completed in 26.61 seconds with avg training loss 306.06\n",
      "epoch 2118 completed in 26.34 seconds with avg training loss 306.2\n",
      "epoch 2119 completed in 26.72 seconds with avg training loss 306.21\n",
      "epoch 2120 completed in 26.64 seconds with avg training loss 306.23\n",
      "model not saved. validation loss was  804.0691406249999\n",
      "epoch 2121 completed in 26.69 seconds with avg training loss 306.49\n",
      "epoch 2122 completed in 26.5 seconds with avg training loss 305.78\n",
      "epoch 2123 completed in 26.67 seconds with avg training loss 306.62\n",
      "epoch 2124 completed in 26.91 seconds with avg training loss 305.82\n",
      "epoch 2125 completed in 27.05 seconds with avg training loss 306.07\n",
      "model not saved. validation loss was  808.1442382812501\n",
      "epoch 2126 completed in 26.94 seconds with avg training loss 306.31\n",
      "epoch 2127 completed in 26.45 seconds with avg training loss 306.54\n",
      "epoch 2128 completed in 26.61 seconds with avg training loss 306.25\n",
      "epoch 2129 completed in 26.52 seconds with avg training loss 306.38\n",
      "epoch 2130 completed in 26.69 seconds with avg training loss 305.92\n",
      "model not saved. validation loss was  807.07998046875\n",
      "epoch 2131 completed in 26.69 seconds with avg training loss 305.98\n",
      "epoch 2132 completed in 26.94 seconds with avg training loss 306.09\n",
      "epoch 2133 completed in 27.23 seconds with avg training loss 305.97\n",
      "epoch 2134 completed in 27.27 seconds with avg training loss 306.4\n",
      "epoch 2135 completed in 26.17 seconds with avg training loss 306.31\n",
      "model not saved. validation loss was  805.1961523437501\n",
      "epoch 2136 completed in 26.63 seconds with avg training loss 306.46\n",
      "epoch 2137 completed in 26.77 seconds with avg training loss 306.53\n",
      "epoch 2138 completed in 26.88 seconds with avg training loss 306.79\n",
      "epoch 2139 completed in 27.09 seconds with avg training loss 306.52\n",
      "epoch 2140 completed in 27.09 seconds with avg training loss 305.73\n",
      "model not saved. validation loss was  811.3411328125\n",
      "epoch 2141 completed in 26.89 seconds with avg training loss 306.25\n",
      "epoch 2142 completed in 27.36 seconds with avg training loss 306.51\n",
      "epoch 2143 completed in 27.27 seconds with avg training loss 306.27\n",
      "epoch 2144 completed in 26.88 seconds with avg training loss 306.39\n",
      "epoch 2145 completed in 27.05 seconds with avg training loss 306.38\n",
      "model not saved. validation loss was  808.4109375\n",
      "epoch 2146 completed in 26.63 seconds with avg training loss 306.05\n",
      "epoch 2147 completed in 27.09 seconds with avg training loss 306.25\n",
      "epoch 2148 completed in 27.45 seconds with avg training loss 306.22\n",
      "epoch 2149 completed in 27.03 seconds with avg training loss 306.17\n",
      "epoch 2150 completed in 27.2 seconds with avg training loss 305.69\n",
      "model not saved. validation loss was  806.56166015625\n",
      "epoch 2151 completed in 26.98 seconds with avg training loss 305.82\n",
      "epoch 2152 completed in 26.47 seconds with avg training loss 306.83\n",
      "epoch 2153 completed in 26.59 seconds with avg training loss 305.62\n",
      "epoch 2154 completed in 26.66 seconds with avg training loss 306.25\n",
      "epoch 2155 completed in 27.06 seconds with avg training loss 306.28\n",
      "model not saved. validation loss was  804.85912109375\n",
      "epoch 2156 completed in 27.34 seconds with avg training loss 306.09\n",
      "epoch 2157 completed in 26.92 seconds with avg training loss 306.31\n",
      "epoch 2158 completed in 27.02 seconds with avg training loss 305.82\n",
      "epoch 2159 completed in 27.03 seconds with avg training loss 306.36\n",
      "epoch 2160 completed in 26.97 seconds with avg training loss 306.24\n",
      "model not saved. validation loss was  807.6901953125\n",
      "epoch 2161 completed in 27.33 seconds with avg training loss 306.51\n",
      "epoch 2162 completed in 27.0 seconds with avg training loss 306.38\n",
      "epoch 2163 completed in 26.81 seconds with avg training loss 306.39\n",
      "epoch 2164 completed in 27.17 seconds with avg training loss 306.17\n",
      "epoch 2165 completed in 26.89 seconds with avg training loss 306.68\n",
      "model not saved. validation loss was  803.3922265625\n",
      "epoch 2166 completed in 26.84 seconds with avg training loss 306.47\n",
      "epoch 2167 completed in 26.69 seconds with avg training loss 306.32\n",
      "epoch 2168 completed in 27.08 seconds with avg training loss 305.83\n",
      "epoch 2169 completed in 27.14 seconds with avg training loss 306.15\n",
      "epoch 2170 completed in 26.67 seconds with avg training loss 306.64\n",
      "model not saved. validation loss was  804.78529296875\n",
      "epoch 2171 completed in 26.63 seconds with avg training loss 306.4\n",
      "epoch 2172 completed in 27.11 seconds with avg training loss 305.64\n",
      "epoch 2173 completed in 26.8 seconds with avg training loss 306.12\n",
      "epoch 2174 completed in 26.7 seconds with avg training loss 306.06\n",
      "epoch 2175 completed in 26.94 seconds with avg training loss 306.44\n",
      "model not saved. validation loss was  801.7287890625\n",
      "epoch 2176 completed in 27.14 seconds with avg training loss 306.19\n",
      "epoch 2177 completed in 26.92 seconds with avg training loss 306.34\n",
      "epoch 2178 completed in 26.72 seconds with avg training loss 306.47\n",
      "epoch 2179 completed in 26.16 seconds with avg training loss 306.89\n",
      "epoch 2180 completed in 26.5 seconds with avg training loss 305.89\n",
      "model not saved. validation loss was  803.41921875\n",
      "epoch 2181 completed in 26.73 seconds with avg training loss 306.31\n",
      "epoch 2182 completed in 26.41 seconds with avg training loss 305.74\n",
      "epoch 2183 completed in 27.47 seconds with avg training loss 306.31\n",
      "epoch 2184 completed in 27.19 seconds with avg training loss 306.6\n",
      "epoch 2185 completed in 27.02 seconds with avg training loss 306.4\n",
      "model not saved. validation loss was  806.9229296875001\n",
      "epoch 2186 completed in 26.69 seconds with avg training loss 306.1\n",
      "epoch 2187 completed in 27.06 seconds with avg training loss 306.45\n",
      "epoch 2188 completed in 26.64 seconds with avg training loss 306.13\n",
      "epoch 2189 completed in 27.03 seconds with avg training loss 306.24\n",
      "epoch 2190 completed in 26.95 seconds with avg training loss 306.66\n",
      "model not saved. validation loss was  806.3372070312499\n",
      "epoch 2191 completed in 27.03 seconds with avg training loss 306.66\n",
      "epoch 2192 completed in 26.48 seconds with avg training loss 306.46\n",
      "epoch 2193 completed in 26.97 seconds with avg training loss 306.37\n",
      "epoch 2194 completed in 26.77 seconds with avg training loss 306.23\n",
      "epoch 2195 completed in 27.06 seconds with avg training loss 306.35\n",
      "model not saved. validation loss was  804.0047851562499\n",
      "epoch 2196 completed in 27.06 seconds with avg training loss 306.22\n",
      "epoch 2197 completed in 26.84 seconds with avg training loss 306.15\n",
      "epoch 2198 completed in 26.86 seconds with avg training loss 306.9\n",
      "epoch 2199 completed in 26.59 seconds with avg training loss 306.89\n",
      "epoch 2200 completed in 26.8 seconds with avg training loss 306.42\n",
      "model not saved. validation loss was  808.55970703125\n",
      "epoch 2201 completed in 26.98 seconds with avg training loss 306.17\n",
      "epoch 2202 completed in 26.45 seconds with avg training loss 306.23\n",
      "epoch 2203 completed in 27.05 seconds with avg training loss 306.22\n",
      "epoch 2204 completed in 26.94 seconds with avg training loss 306.17\n",
      "epoch 2205 completed in 26.48 seconds with avg training loss 305.96\n",
      "model not saved. validation loss was  810.7541015625\n",
      "epoch 2206 completed in 26.97 seconds with avg training loss 305.88\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 2207 completed in 26.84 seconds with avg training loss 306.31\n",
      "epoch 2208 completed in 27.06 seconds with avg training loss 305.94\n",
      "epoch 2209 completed in 27.03 seconds with avg training loss 306.47\n",
      "epoch 2210 completed in 26.94 seconds with avg training loss 306.63\n",
      "model not saved. validation loss was  807.08841796875\n",
      "epoch 2211 completed in 27.08 seconds with avg training loss 306.52\n",
      "epoch 2212 completed in 26.92 seconds with avg training loss 306.36\n",
      "epoch 2213 completed in 26.31 seconds with avg training loss 306.45\n",
      "epoch 2214 completed in 26.7 seconds with avg training loss 305.71\n",
      "epoch 2215 completed in 26.33 seconds with avg training loss 306.33\n",
      "model not saved. validation loss was  801.02470703125\n",
      "epoch 2216 completed in 26.42 seconds with avg training loss 306.29\n",
      "epoch 2217 completed in 27.0 seconds with avg training loss 306.38\n",
      "epoch 2218 completed in 26.61 seconds with avg training loss 306.26\n",
      "epoch 2219 completed in 26.47 seconds with avg training loss 306.46\n",
      "epoch 2220 completed in 26.91 seconds with avg training loss 306.74\n",
      "model not saved. validation loss was  798.6136132812501\n",
      "epoch 2221 completed in 26.83 seconds with avg training loss 305.86\n",
      "epoch 2222 completed in 26.59 seconds with avg training loss 306.04\n",
      "epoch 2223 completed in 26.44 seconds with avg training loss 306.39\n",
      "epoch 2224 completed in 26.92 seconds with avg training loss 305.97\n",
      "epoch 2225 completed in 27.09 seconds with avg training loss 306.68\n",
      "model not saved. validation loss was  803.38384765625\n",
      "epoch 2226 completed in 26.97 seconds with avg training loss 306.7\n",
      "epoch 2227 completed in 26.73 seconds with avg training loss 306.23\n",
      "epoch 2228 completed in 27.17 seconds with avg training loss 306.79\n",
      "epoch 2229 completed in 26.7 seconds with avg training loss 305.96\n",
      "epoch 2230 completed in 27.23 seconds with avg training loss 306.21\n",
      "model not saved. validation loss was  808.75775390625\n",
      "epoch 2231 completed in 26.91 seconds with avg training loss 306.16\n",
      "epoch 2232 completed in 26.92 seconds with avg training loss 306.41\n",
      "epoch 2233 completed in 27.11 seconds with avg training loss 305.51\n",
      "epoch 2234 completed in 27.23 seconds with avg training loss 306.24\n",
      "epoch 2235 completed in 27.41 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  800.2334765625001\n",
      "epoch 2236 completed in 27.03 seconds with avg training loss 306.19\n",
      "epoch 2237 completed in 27.13 seconds with avg training loss 306.13\n",
      "epoch 2238 completed in 26.88 seconds with avg training loss 306.17\n",
      "epoch 2239 completed in 27.2 seconds with avg training loss 306.56\n",
      "epoch 2240 completed in 26.7 seconds with avg training loss 306.39\n",
      "model not saved. validation loss was  813.89357421875\n",
      "epoch 2241 completed in 27.31 seconds with avg training loss 306.45\n",
      "epoch 2242 completed in 26.84 seconds with avg training loss 306.29\n",
      "epoch 2243 completed in 26.95 seconds with avg training loss 306.45\n",
      "epoch 2244 completed in 27.14 seconds with avg training loss 306.3\n",
      "epoch 2245 completed in 26.91 seconds with avg training loss 306.15\n",
      "model not saved. validation loss was  805.3390625\n",
      "epoch 2246 completed in 27.48 seconds with avg training loss 306.22\n",
      "epoch 2247 completed in 26.66 seconds with avg training loss 306.04\n",
      "epoch 2248 completed in 26.5 seconds with avg training loss 306.4\n",
      "epoch 2249 completed in 26.41 seconds with avg training loss 306.16\n",
      "epoch 2250 completed in 26.64 seconds with avg training loss 306.38\n",
      "model not saved. validation loss was  804.55224609375\n",
      "epoch 2251 completed in 26.39 seconds with avg training loss 306.35\n",
      "epoch 2252 completed in 26.81 seconds with avg training loss 306.33\n",
      "epoch 2253 completed in 26.56 seconds with avg training loss 306.02\n",
      "epoch 2254 completed in 26.91 seconds with avg training loss 306.51\n",
      "epoch 2255 completed in 26.73 seconds with avg training loss 306.48\n",
      "model not saved. validation loss was  806.431328125\n",
      "epoch 2256 completed in 26.33 seconds with avg training loss 305.58\n",
      "epoch 2257 completed in 26.89 seconds with avg training loss 306.53\n",
      "epoch 2258 completed in 26.22 seconds with avg training loss 306.68\n",
      "epoch 2259 completed in 26.58 seconds with avg training loss 305.77\n",
      "epoch 2260 completed in 26.97 seconds with avg training loss 306.13\n",
      "model not saved. validation loss was  804.27072265625\n",
      "epoch 2261 completed in 26.78 seconds with avg training loss 306.49\n",
      "epoch 2262 completed in 26.94 seconds with avg training loss 305.79\n",
      "epoch 2263 completed in 26.5 seconds with avg training loss 306.53\n",
      "epoch 2264 completed in 26.88 seconds with avg training loss 306.47\n",
      "epoch 2265 completed in 27.23 seconds with avg training loss 305.71\n",
      "model not saved. validation loss was  809.951796875\n",
      "epoch 2266 completed in 26.91 seconds with avg training loss 305.8\n",
      "epoch 2267 completed in 26.61 seconds with avg training loss 306.63\n",
      "epoch 2268 completed in 27.03 seconds with avg training loss 306.24\n",
      "epoch 2269 completed in 27.03 seconds with avg training loss 305.81\n",
      "epoch 2270 completed in 26.95 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  801.1299804687501\n",
      "epoch 2271 completed in 26.83 seconds with avg training loss 305.8\n",
      "epoch 2272 completed in 27.2 seconds with avg training loss 306.38\n",
      "epoch 2273 completed in 27.09 seconds with avg training loss 305.85\n",
      "epoch 2274 completed in 27.03 seconds with avg training loss 306.55\n",
      "epoch 2275 completed in 27.25 seconds with avg training loss 305.93\n",
      "model not saved. validation loss was  804.1734375000001\n",
      "epoch 2276 completed in 26.98 seconds with avg training loss 306.1\n",
      "epoch 2277 completed in 26.89 seconds with avg training loss 306.22\n",
      "epoch 2278 completed in 26.91 seconds with avg training loss 306.27\n",
      "epoch 2279 completed in 26.59 seconds with avg training loss 305.98\n",
      "epoch 2280 completed in 26.7 seconds with avg training loss 305.97\n",
      "model not saved. validation loss was  801.6845312500001\n",
      "epoch 2281 completed in 25.73 seconds with avg training loss 306.07\n",
      "epoch 2282 completed in 27.06 seconds with avg training loss 305.89\n",
      "epoch 2283 completed in 26.5 seconds with avg training loss 305.98\n",
      "epoch 2284 completed in 26.44 seconds with avg training loss 305.86\n",
      "epoch 2285 completed in 26.81 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  810.66787109375\n",
      "epoch 2286 completed in 26.63 seconds with avg training loss 306.03\n",
      "epoch 2287 completed in 26.8 seconds with avg training loss 306.3\n",
      "epoch 2288 completed in 26.84 seconds with avg training loss 306.17\n",
      "epoch 2289 completed in 26.89 seconds with avg training loss 306.13\n",
      "epoch 2290 completed in 26.27 seconds with avg training loss 306.35\n",
      "model not saved. validation loss was  803.4318164062499\n",
      "epoch 2291 completed in 26.67 seconds with avg training loss 306.26\n",
      "epoch 2292 completed in 27.09 seconds with avg training loss 305.72\n",
      "epoch 2293 completed in 27.06 seconds with avg training loss 306.15\n",
      "epoch 2294 completed in 27.2 seconds with avg training loss 305.94\n",
      "epoch 2295 completed in 26.64 seconds with avg training loss 305.95\n",
      "model not saved. validation loss was  794.99646484375\n",
      "epoch 2296 completed in 26.72 seconds with avg training loss 306.28\n",
      "epoch 2297 completed in 26.61 seconds with avg training loss 306.22\n",
      "epoch 2298 completed in 26.38 seconds with avg training loss 305.69\n",
      "epoch 2299 completed in 27.17 seconds with avg training loss 306.02\n",
      "epoch 2300 completed in 26.78 seconds with avg training loss 306.09\n",
      "model not saved. validation loss was  809.9975390625\n",
      "epoch 2301 completed in 26.16 seconds with avg training loss 306.3\n",
      "epoch 2302 completed in 26.88 seconds with avg training loss 306.31\n",
      "epoch 2303 completed in 26.53 seconds with avg training loss 306.22\n",
      "epoch 2304 completed in 26.91 seconds with avg training loss 305.77\n",
      "epoch 2305 completed in 27.09 seconds with avg training loss 305.91\n",
      "model not saved. validation loss was  802.20625\n",
      "epoch 2306 completed in 26.44 seconds with avg training loss 305.94\n",
      "epoch 2307 completed in 27.02 seconds with avg training loss 306.35\n",
      "epoch 2308 completed in 26.83 seconds with avg training loss 306.14\n",
      "epoch 2309 completed in 26.84 seconds with avg training loss 305.78\n",
      "epoch 2310 completed in 26.84 seconds with avg training loss 307.01\n",
      "model not saved. validation loss was  801.1481054687499\n",
      "epoch 2311 completed in 26.89 seconds with avg training loss 306.62\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 2312 completed in 26.91 seconds with avg training loss 306.64\n",
      "epoch 2313 completed in 27.0 seconds with avg training loss 306.11\n",
      "epoch 2314 completed in 26.56 seconds with avg training loss 306.21\n",
      "epoch 2315 completed in 26.98 seconds with avg training loss 306.7\n",
      "model not saved. validation loss was  813.605390625\n",
      "epoch 2316 completed in 26.59 seconds with avg training loss 306.35\n",
      "epoch 2317 completed in 26.13 seconds with avg training loss 306.23\n",
      "epoch 2318 completed in 27.08 seconds with avg training loss 305.93\n",
      "epoch 2319 completed in 26.61 seconds with avg training loss 306.06\n",
      "epoch 2320 completed in 27.02 seconds with avg training loss 306.35\n",
      "model not saved. validation loss was  811.74888671875\n",
      "epoch 2321 completed in 26.86 seconds with avg training loss 306.17\n",
      "epoch 2322 completed in 27.11 seconds with avg training loss 306.42\n",
      "epoch 2323 completed in 26.92 seconds with avg training loss 306.1\n",
      "epoch 2324 completed in 26.75 seconds with avg training loss 306.11\n",
      "epoch 2325 completed in 27.34 seconds with avg training loss 306.37\n",
      "model not saved. validation loss was  810.17796875\n",
      "epoch 2326 completed in 26.81 seconds with avg training loss 305.95\n",
      "epoch 2327 completed in 27.11 seconds with avg training loss 306.32\n",
      "epoch 2328 completed in 26.89 seconds with avg training loss 305.81\n",
      "epoch 2329 completed in 26.84 seconds with avg training loss 306.12\n",
      "epoch 2330 completed in 27.25 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  805.16208984375\n",
      "epoch 2331 completed in 26.95 seconds with avg training loss 306.51\n",
      "epoch 2332 completed in 26.63 seconds with avg training loss 306.4\n",
      "epoch 2333 completed in 26.67 seconds with avg training loss 306.81\n",
      "epoch 2334 completed in 26.7 seconds with avg training loss 305.96\n",
      "epoch 2335 completed in 26.75 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  800.8979296875\n",
      "epoch 2336 completed in 26.94 seconds with avg training loss 306.57\n",
      "epoch 2337 completed in 27.06 seconds with avg training loss 306.55\n",
      "epoch 2338 completed in 27.2 seconds with avg training loss 306.48\n",
      "epoch 2339 completed in 27.05 seconds with avg training loss 306.54\n",
      "epoch 2340 completed in 26.86 seconds with avg training loss 306.36\n",
      "model not saved. validation loss was  798.92580078125\n",
      "epoch 2341 completed in 27.03 seconds with avg training loss 306.52\n",
      "epoch 2342 completed in 26.86 seconds with avg training loss 306.17\n",
      "epoch 2343 completed in 26.98 seconds with avg training loss 306.52\n",
      "epoch 2344 completed in 27.05 seconds with avg training loss 306.13\n",
      "epoch 2345 completed in 27.27 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  803.19294921875\n",
      "epoch 2346 completed in 26.73 seconds with avg training loss 306.15\n",
      "epoch 2347 completed in 26.58 seconds with avg training loss 305.73\n",
      "epoch 2348 completed in 26.64 seconds with avg training loss 305.55\n",
      "epoch 2349 completed in 26.78 seconds with avg training loss 306.67\n",
      "epoch 2350 completed in 26.59 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  808.6037890625\n",
      "epoch 2351 completed in 26.84 seconds with avg training loss 306.34\n",
      "epoch 2352 completed in 26.92 seconds with avg training loss 306.42\n",
      "epoch 2353 completed in 26.36 seconds with avg training loss 306.28\n",
      "epoch 2354 completed in 26.56 seconds with avg training loss 306.53\n",
      "epoch 2355 completed in 26.83 seconds with avg training loss 306.16\n",
      "model not saved. validation loss was  808.637734375\n",
      "epoch 2356 completed in 26.45 seconds with avg training loss 306.13\n",
      "epoch 2357 completed in 26.86 seconds with avg training loss 306.6\n",
      "epoch 2358 completed in 26.83 seconds with avg training loss 305.84\n",
      "epoch 2359 completed in 26.45 seconds with avg training loss 305.69\n",
      "epoch 2360 completed in 26.78 seconds with avg training loss 305.43\n",
      "model not saved. validation loss was  799.46857421875\n",
      "epoch 2361 completed in 26.67 seconds with avg training loss 305.47\n",
      "epoch 2362 completed in 27.13 seconds with avg training loss 305.93\n",
      "epoch 2363 completed in 27.06 seconds with avg training loss 305.76\n",
      "epoch 2364 completed in 27.14 seconds with avg training loss 306.34\n",
      "epoch 2365 completed in 26.91 seconds with avg training loss 305.97\n",
      "model not saved. validation loss was  801.1668164062501\n",
      "epoch 2366 completed in 26.84 seconds with avg training loss 306.26\n",
      "epoch 2367 completed in 26.5 seconds with avg training loss 305.85\n",
      "epoch 2368 completed in 26.39 seconds with avg training loss 306.75\n",
      "epoch 2369 completed in 26.98 seconds with avg training loss 306.04\n",
      "epoch 2370 completed in 26.88 seconds with avg training loss 305.58\n",
      "model not saved. validation loss was  799.938359375\n",
      "epoch 2371 completed in 26.55 seconds with avg training loss 305.85\n",
      "epoch 2372 completed in 26.86 seconds with avg training loss 306.79\n",
      "epoch 2373 completed in 27.14 seconds with avg training loss 306.02\n",
      "epoch 2374 completed in 26.72 seconds with avg training loss 306.02\n",
      "epoch 2375 completed in 26.48 seconds with avg training loss 306.39\n",
      "model not saved. validation loss was  805.3009374999999\n",
      "epoch 2376 completed in 26.91 seconds with avg training loss 306.35\n",
      "epoch 2377 completed in 26.94 seconds with avg training loss 306.56\n",
      "epoch 2378 completed in 26.94 seconds with avg training loss 306.15\n",
      "epoch 2379 completed in 26.53 seconds with avg training loss 306.01\n",
      "epoch 2380 completed in 26.56 seconds with avg training loss 306.5\n",
      "model not saved. validation loss was  811.3096484375001\n",
      "epoch 2381 completed in 26.97 seconds with avg training loss 306.3\n",
      "epoch 2382 completed in 26.94 seconds with avg training loss 306.01\n",
      "epoch 2383 completed in 27.14 seconds with avg training loss 305.92\n",
      "epoch 2384 completed in 27.09 seconds with avg training loss 306.09\n",
      "epoch 2385 completed in 27.36 seconds with avg training loss 306.35\n",
      "model not saved. validation loss was  799.09244140625\n",
      "epoch 2386 completed in 26.73 seconds with avg training loss 306.03\n",
      "epoch 2387 completed in 27.14 seconds with avg training loss 306.26\n",
      "epoch 2388 completed in 26.94 seconds with avg training loss 305.78\n",
      "epoch 2389 completed in 26.8 seconds with avg training loss 306.02\n",
      "epoch 2390 completed in 27.19 seconds with avg training loss 306.09\n",
      "model not saved. validation loss was  804.8014453124999\n",
      "epoch 2391 completed in 27.0 seconds with avg training loss 305.97\n",
      "epoch 2392 completed in 27.33 seconds with avg training loss 306.08\n",
      "epoch 2393 completed in 27.2 seconds with avg training loss 305.82\n",
      "epoch 2394 completed in 26.88 seconds with avg training loss 306.28\n",
      "epoch 2395 completed in 27.05 seconds with avg training loss 306.1\n",
      "model not saved. validation loss was  809.2260546875001\n",
      "epoch 2396 completed in 27.17 seconds with avg training loss 306.43\n",
      "epoch 2397 completed in 26.83 seconds with avg training loss 306.31\n",
      "epoch 2398 completed in 27.69 seconds with avg training loss 306.38\n",
      "epoch 2399 completed in 27.2 seconds with avg training loss 306.3\n",
      "epoch 2400 completed in 26.69 seconds with avg training loss 305.98\n",
      "model not saved. validation loss was  805.36015625\n",
      "epoch 2401 completed in 27.89 seconds with avg training loss 306.0\n",
      "epoch 2402 completed in 27.23 seconds with avg training loss 306.08\n",
      "epoch 2403 completed in 26.86 seconds with avg training loss 306.62\n",
      "epoch 2404 completed in 27.03 seconds with avg training loss 306.34\n",
      "epoch 2405 completed in 26.95 seconds with avg training loss 306.17\n",
      "model not saved. validation loss was  805.56765625\n",
      "epoch 2406 completed in 27.13 seconds with avg training loss 306.12\n",
      "epoch 2407 completed in 27.34 seconds with avg training loss 306.1\n",
      "epoch 2408 completed in 26.95 seconds with avg training loss 306.47\n",
      "epoch 2409 completed in 26.27 seconds with avg training loss 306.72\n",
      "epoch 2410 completed in 26.64 seconds with avg training loss 306.34\n",
      "model not saved. validation loss was  802.62359375\n",
      "epoch 2411 completed in 26.31 seconds with avg training loss 306.49\n",
      "epoch 2412 completed in 26.91 seconds with avg training loss 306.28\n",
      "epoch 2413 completed in 26.64 seconds with avg training loss 305.92\n",
      "epoch 2414 completed in 26.52 seconds with avg training loss 305.81\n",
      "epoch 2415 completed in 26.56 seconds with avg training loss 306.17\n",
      "model not saved. validation loss was  803.881328125\n",
      "epoch 2416 completed in 27.03 seconds with avg training loss 306.18\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 2417 completed in 27.11 seconds with avg training loss 306.3\n",
      "epoch 2418 completed in 27.19 seconds with avg training loss 306.5\n",
      "epoch 2419 completed in 26.81 seconds with avg training loss 306.17\n",
      "epoch 2420 completed in 27.06 seconds with avg training loss 306.69\n",
      "model not saved. validation loss was  802.2812109375\n",
      "epoch 2421 completed in 27.27 seconds with avg training loss 305.92\n",
      "epoch 2422 completed in 27.28 seconds with avg training loss 306.2\n",
      "epoch 2423 completed in 27.28 seconds with avg training loss 306.46\n",
      "epoch 2424 completed in 27.11 seconds with avg training loss 306.48\n",
      "epoch 2425 completed in 26.91 seconds with avg training loss 306.15\n",
      "model not saved. validation loss was  812.7708984375\n",
      "epoch 2426 completed in 27.16 seconds with avg training loss 305.88\n",
      "epoch 2427 completed in 26.72 seconds with avg training loss 306.56\n",
      "epoch 2428 completed in 27.25 seconds with avg training loss 306.05\n",
      "epoch 2429 completed in 27.14 seconds with avg training loss 306.22\n",
      "epoch 2430 completed in 26.95 seconds with avg training loss 306.03\n",
      "model not saved. validation loss was  803.9843164062501\n",
      "epoch 2431 completed in 26.7 seconds with avg training loss 306.35\n",
      "epoch 2432 completed in 26.73 seconds with avg training loss 306.1\n",
      "epoch 2433 completed in 26.77 seconds with avg training loss 306.53\n",
      "epoch 2434 completed in 26.58 seconds with avg training loss 305.99\n",
      "epoch 2435 completed in 27.41 seconds with avg training loss 306.54\n",
      "model not saved. validation loss was  804.513203125\n",
      "epoch 2436 completed in 26.83 seconds with avg training loss 306.5\n",
      "epoch 2437 completed in 27.13 seconds with avg training loss 306.55\n",
      "epoch 2438 completed in 27.23 seconds with avg training loss 306.73\n",
      "epoch 2439 completed in 27.36 seconds with avg training loss 306.09\n",
      "epoch 2440 completed in 27.34 seconds with avg training loss 305.97\n",
      "model not saved. validation loss was  813.0820703124999\n",
      "epoch 2441 completed in 27.72 seconds with avg training loss 306.46\n",
      "epoch 2442 completed in 27.27 seconds with avg training loss 306.19\n",
      "epoch 2443 completed in 26.89 seconds with avg training loss 305.67\n",
      "epoch 2444 completed in 26.66 seconds with avg training loss 305.93\n",
      "epoch 2445 completed in 26.69 seconds with avg training loss 306.49\n",
      "model not saved. validation loss was  800.1134375\n",
      "epoch 2446 completed in 27.06 seconds with avg training loss 306.24\n",
      "epoch 2447 completed in 26.59 seconds with avg training loss 306.23\n",
      "epoch 2448 completed in 26.81 seconds with avg training loss 305.93\n",
      "epoch 2449 completed in 26.67 seconds with avg training loss 306.44\n",
      "epoch 2450 completed in 26.44 seconds with avg training loss 306.93\n",
      "model not saved. validation loss was  807.3383984375\n",
      "epoch 2451 completed in 26.89 seconds with avg training loss 306.27\n",
      "epoch 2452 completed in 26.83 seconds with avg training loss 306.44\n",
      "epoch 2453 completed in 26.95 seconds with avg training loss 305.91\n",
      "epoch 2454 completed in 26.64 seconds with avg training loss 306.1\n",
      "epoch 2455 completed in 26.52 seconds with avg training loss 306.47\n",
      "model not saved. validation loss was  805.98525390625\n",
      "epoch 2456 completed in 26.78 seconds with avg training loss 306.05\n",
      "epoch 2457 completed in 26.66 seconds with avg training loss 306.63\n",
      "epoch 2458 completed in 26.36 seconds with avg training loss 307.18\n",
      "epoch 2459 completed in 26.77 seconds with avg training loss 305.93\n",
      "epoch 2460 completed in 26.61 seconds with avg training loss 305.87\n",
      "model not saved. validation loss was  802.83291015625\n",
      "epoch 2461 completed in 27.22 seconds with avg training loss 305.8\n",
      "epoch 2462 completed in 26.83 seconds with avg training loss 305.1\n",
      "epoch 2463 completed in 26.77 seconds with avg training loss 305.97\n",
      "epoch 2464 completed in 26.64 seconds with avg training loss 306.22\n",
      "epoch 2465 completed in 26.55 seconds with avg training loss 306.44\n",
      "model not saved. validation loss was  809.1698437499999\n",
      "epoch 2466 completed in 26.75 seconds with avg training loss 306.12\n",
      "epoch 2467 completed in 26.97 seconds with avg training loss 306.06\n",
      "epoch 2468 completed in 26.98 seconds with avg training loss 306.27\n",
      "epoch 2469 completed in 26.95 seconds with avg training loss 305.7\n",
      "epoch 2470 completed in 26.83 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  808.0501562500001\n",
      "epoch 2471 completed in 26.92 seconds with avg training loss 306.24\n",
      "epoch 2472 completed in 26.78 seconds with avg training loss 305.94\n",
      "epoch 2473 completed in 26.8 seconds with avg training loss 306.19\n",
      "epoch 2474 completed in 26.58 seconds with avg training loss 306.59\n",
      "epoch 2475 completed in 27.0 seconds with avg training loss 306.56\n",
      "model not saved. validation loss was  807.9125195312499\n",
      "epoch 2476 completed in 27.16 seconds with avg training loss 306.31\n",
      "epoch 2477 completed in 27.13 seconds with avg training loss 306.46\n",
      "epoch 2478 completed in 27.25 seconds with avg training loss 305.55\n",
      "epoch 2479 completed in 26.84 seconds with avg training loss 306.31\n",
      "epoch 2480 completed in 26.47 seconds with avg training loss 305.87\n",
      "model not saved. validation loss was  802.64529296875\n",
      "epoch 2481 completed in 26.48 seconds with avg training loss 306.51\n",
      "epoch 2482 completed in 26.45 seconds with avg training loss 305.9\n",
      "epoch 2483 completed in 26.81 seconds with avg training loss 305.92\n",
      "epoch 2484 completed in 27.13 seconds with avg training loss 306.64\n",
      "epoch 2485 completed in 27.03 seconds with avg training loss 306.25\n",
      "model not saved. validation loss was  804.3037890625\n",
      "epoch 2486 completed in 27.17 seconds with avg training loss 306.0\n",
      "epoch 2487 completed in 26.88 seconds with avg training loss 305.97\n",
      "epoch 2488 completed in 27.0 seconds with avg training loss 305.83\n",
      "epoch 2489 completed in 27.11 seconds with avg training loss 306.07\n",
      "epoch 2490 completed in 27.13 seconds with avg training loss 306.0\n",
      "model not saved. validation loss was  807.92576171875\n",
      "epoch 2491 completed in 26.41 seconds with avg training loss 306.56\n",
      "epoch 2492 completed in 26.75 seconds with avg training loss 305.9\n",
      "epoch 2493 completed in 27.13 seconds with avg training loss 306.4\n",
      "epoch 2494 completed in 27.09 seconds with avg training loss 305.9\n",
      "epoch 2495 completed in 26.81 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  805.2650390625\n",
      "epoch 2496 completed in 27.23 seconds with avg training loss 306.15\n",
      "epoch 2497 completed in 27.11 seconds with avg training loss 306.03\n",
      "epoch 2498 completed in 27.22 seconds with avg training loss 306.28\n",
      "epoch 2499 completed in 27.09 seconds with avg training loss 306.37\n",
      "epoch 2500 completed in 27.0 seconds with avg training loss 306.22\n",
      "model not saved. validation loss was  810.34810546875\n",
      "epoch 2501 completed in 27.56 seconds with avg training loss 306.07\n",
      "epoch 2502 completed in 26.78 seconds with avg training loss 306.34\n",
      "epoch 2503 completed in 26.7 seconds with avg training loss 306.55\n",
      "epoch 2504 completed in 26.77 seconds with avg training loss 305.95\n",
      "epoch 2505 completed in 26.89 seconds with avg training loss 305.67\n",
      "model not saved. validation loss was  804.4039843749999\n",
      "epoch 2506 completed in 27.14 seconds with avg training loss 306.03\n",
      "epoch 2507 completed in 26.75 seconds with avg training loss 306.45\n",
      "epoch 2508 completed in 26.84 seconds with avg training loss 306.46\n",
      "epoch 2509 completed in 27.09 seconds with avg training loss 306.6\n",
      "epoch 2510 completed in 26.88 seconds with avg training loss 306.11\n",
      "model not saved. validation loss was  806.1339257812499\n",
      "epoch 2511 completed in 26.89 seconds with avg training loss 306.0\n",
      "epoch 2512 completed in 43.77 seconds with avg training loss 305.86\n",
      "epoch 2513 completed in 170.81 seconds with avg training loss 305.35\n",
      "epoch 2514 completed in 26.48 seconds with avg training loss 306.22\n",
      "epoch 2515 completed in 27.11 seconds with avg training loss 306.16\n",
      "model not saved. validation loss was  803.2640039062501\n",
      "epoch 2516 completed in 27.05 seconds with avg training loss 306.31\n",
      "epoch 2517 completed in 26.8 seconds with avg training loss 306.11\n",
      "epoch 2518 completed in 26.8 seconds with avg training loss 306.48\n",
      "epoch 2519 completed in 26.8 seconds with avg training loss 306.63\n",
      "epoch 2520 completed in 26.78 seconds with avg training loss 305.94\n",
      "model not saved. validation loss was  813.5157812500001\n",
      "epoch 2521 completed in 27.2 seconds with avg training loss 306.35\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 2522 completed in 26.95 seconds with avg training loss 306.28\n",
      "epoch 2523 completed in 27.08 seconds with avg training loss 305.37\n",
      "epoch 2524 completed in 26.95 seconds with avg training loss 306.23\n",
      "epoch 2525 completed in 27.16 seconds with avg training loss 305.83\n",
      "model not saved. validation loss was  805.11572265625\n",
      "epoch 2526 completed in 27.0 seconds with avg training loss 306.32\n",
      "epoch 2527 completed in 27.05 seconds with avg training loss 305.96\n",
      "epoch 2528 completed in 27.0 seconds with avg training loss 305.79\n",
      "epoch 2529 completed in 27.14 seconds with avg training loss 306.15\n",
      "epoch 2530 completed in 27.08 seconds with avg training loss 306.38\n",
      "model not saved. validation loss was  806.80740234375\n",
      "epoch 2531 completed in 27.0 seconds with avg training loss 306.59\n",
      "epoch 2532 completed in 27.28 seconds with avg training loss 306.41\n",
      "epoch 2533 completed in 26.86 seconds with avg training loss 305.65\n",
      "epoch 2534 completed in 27.31 seconds with avg training loss 305.97\n",
      "epoch 2535 completed in 27.13 seconds with avg training loss 305.88\n",
      "model not saved. validation loss was  807.5413867187501\n",
      "epoch 2536 completed in 27.08 seconds with avg training loss 306.57\n",
      "epoch 2537 completed in 27.5 seconds with avg training loss 306.21\n",
      "epoch 2538 completed in 27.06 seconds with avg training loss 306.55\n",
      "epoch 2539 completed in 26.97 seconds with avg training loss 306.76\n",
      "epoch 2540 completed in 27.09 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  803.0269921874999\n",
      "epoch 2541 completed in 26.94 seconds with avg training loss 305.67\n",
      "epoch 2542 completed in 27.36 seconds with avg training loss 306.15\n",
      "epoch 2543 completed in 27.36 seconds with avg training loss 306.39\n",
      "epoch 2544 completed in 26.55 seconds with avg training loss 306.22\n",
      "epoch 2545 completed in 27.19 seconds with avg training loss 306.08\n",
      "model not saved. validation loss was  805.9280859375001\n",
      "epoch 2546 completed in 27.34 seconds with avg training loss 305.86\n",
      "epoch 2547 completed in 27.08 seconds with avg training loss 306.08\n",
      "epoch 2548 completed in 26.94 seconds with avg training loss 305.68\n",
      "epoch 2549 completed in 26.83 seconds with avg training loss 306.06\n",
      "epoch 2550 completed in 27.27 seconds with avg training loss 305.99\n",
      "model not saved. validation loss was  811.1218945312501\n",
      "epoch 2551 completed in 26.84 seconds with avg training loss 306.41\n",
      "epoch 2552 completed in 27.06 seconds with avg training loss 306.0\n",
      "epoch 2553 completed in 27.09 seconds with avg training loss 305.69\n",
      "epoch 2554 completed in 27.03 seconds with avg training loss 306.15\n",
      "epoch 2555 completed in 26.94 seconds with avg training loss 305.89\n",
      "model not saved. validation loss was  808.2333203125\n",
      "epoch 2556 completed in 26.66 seconds with avg training loss 306.14\n",
      "epoch 2557 completed in 26.73 seconds with avg training loss 306.26\n",
      "epoch 2558 completed in 27.14 seconds with avg training loss 306.32\n",
      "epoch 2559 completed in 26.83 seconds with avg training loss 306.19\n",
      "epoch 2560 completed in 26.89 seconds with avg training loss 305.81\n",
      "model not saved. validation loss was  809.7159570312499\n",
      "epoch 2561 completed in 26.8 seconds with avg training loss 306.37\n",
      "epoch 2562 completed in 26.84 seconds with avg training loss 306.21\n",
      "epoch 2563 completed in 27.06 seconds with avg training loss 306.16\n",
      "epoch 2564 completed in 27.09 seconds with avg training loss 306.25\n",
      "epoch 2565 completed in 27.23 seconds with avg training loss 306.27\n",
      "model not saved. validation loss was  807.3408984375001\n",
      "epoch 2566 completed in 27.19 seconds with avg training loss 305.81\n",
      "epoch 2567 completed in 27.45 seconds with avg training loss 306.03\n",
      "epoch 2568 completed in 26.91 seconds with avg training loss 306.36\n",
      "epoch 2569 completed in 27.16 seconds with avg training loss 306.21\n",
      "epoch 2570 completed in 27.34 seconds with avg training loss 306.25\n",
      "model not saved. validation loss was  804.2071874999999\n",
      "epoch 2571 completed in 27.09 seconds with avg training loss 305.96\n",
      "epoch 2572 completed in 27.0 seconds with avg training loss 306.37\n",
      "epoch 2573 completed in 27.14 seconds with avg training loss 306.01\n",
      "epoch 2574 completed in 27.0 seconds with avg training loss 306.59\n",
      "epoch 2575 completed in 26.72 seconds with avg training loss 306.0\n",
      "model not saved. validation loss was  805.4562890625\n",
      "epoch 2576 completed in 27.02 seconds with avg training loss 306.48\n",
      "epoch 2577 completed in 27.22 seconds with avg training loss 306.26\n",
      "epoch 2578 completed in 26.8 seconds with avg training loss 306.12\n",
      "epoch 2579 completed in 26.97 seconds with avg training loss 306.09\n",
      "epoch 2580 completed in 26.83 seconds with avg training loss 305.99\n",
      "model not saved. validation loss was  803.93333984375\n",
      "epoch 2581 completed in 27.13 seconds with avg training loss 306.4\n",
      "epoch 2582 completed in 27.25 seconds with avg training loss 305.91\n",
      "epoch 2583 completed in 26.95 seconds with avg training loss 305.36\n",
      "epoch 2584 completed in 27.08 seconds with avg training loss 306.43\n",
      "epoch 2585 completed in 27.14 seconds with avg training loss 306.37\n",
      "model not saved. validation loss was  805.05107421875\n",
      "epoch 2586 completed in 27.0 seconds with avg training loss 306.45\n",
      "epoch 2587 completed in 27.11 seconds with avg training loss 306.33\n",
      "epoch 2588 completed in 27.45 seconds with avg training loss 305.92\n",
      "epoch 2589 completed in 27.27 seconds with avg training loss 306.3\n",
      "epoch 2590 completed in 27.09 seconds with avg training loss 306.36\n",
      "model not saved. validation loss was  804.98505859375\n",
      "epoch 2591 completed in 27.02 seconds with avg training loss 305.75\n",
      "epoch 2592 completed in 26.77 seconds with avg training loss 305.33\n",
      "epoch 2593 completed in 26.7 seconds with avg training loss 305.78\n",
      "epoch 2594 completed in 27.28 seconds with avg training loss 306.56\n",
      "epoch 2595 completed in 27.0 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  801.879609375\n",
      "epoch 2596 completed in 27.06 seconds with avg training loss 306.32\n",
      "epoch 2597 completed in 27.27 seconds with avg training loss 306.29\n",
      "epoch 2598 completed in 26.7 seconds with avg training loss 306.53\n",
      "epoch 2599 completed in 26.84 seconds with avg training loss 306.43\n",
      "epoch 2600 completed in 26.89 seconds with avg training loss 306.13\n",
      "model not saved. validation loss was  805.0162499999999\n",
      "epoch 2601 completed in 27.03 seconds with avg training loss 306.22\n",
      "epoch 2602 completed in 26.78 seconds with avg training loss 306.64\n",
      "epoch 2603 completed in 26.83 seconds with avg training loss 306.41\n",
      "epoch 2604 completed in 26.83 seconds with avg training loss 306.45\n",
      "epoch 2605 completed in 26.81 seconds with avg training loss 306.63\n",
      "model not saved. validation loss was  809.37759765625\n",
      "epoch 2606 completed in 27.11 seconds with avg training loss 306.4\n",
      "epoch 2607 completed in 26.95 seconds with avg training loss 307.0\n",
      "epoch 2608 completed in 27.16 seconds with avg training loss 306.22\n",
      "epoch 2609 completed in 27.03 seconds with avg training loss 306.0\n",
      "epoch 2610 completed in 27.16 seconds with avg training loss 306.41\n",
      "model not saved. validation loss was  815.37298828125\n",
      "epoch 2611 completed in 27.08 seconds with avg training loss 306.02\n",
      "epoch 2612 completed in 27.03 seconds with avg training loss 305.69\n",
      "epoch 2613 completed in 27.13 seconds with avg training loss 306.19\n",
      "epoch 2614 completed in 26.81 seconds with avg training loss 306.5\n",
      "epoch 2615 completed in 26.95 seconds with avg training loss 305.91\n",
      "model not saved. validation loss was  809.29517578125\n",
      "epoch 2616 completed in 27.36 seconds with avg training loss 306.4\n",
      "epoch 2617 completed in 26.67 seconds with avg training loss 306.63\n",
      "epoch 2618 completed in 26.88 seconds with avg training loss 306.21\n",
      "epoch 2619 completed in 26.3 seconds with avg training loss 305.89\n",
      "epoch 2620 completed in 26.89 seconds with avg training loss 306.19\n",
      "model not saved. validation loss was  805.8766015625\n",
      "epoch 2621 completed in 26.89 seconds with avg training loss 305.91\n",
      "epoch 2622 completed in 26.91 seconds with avg training loss 306.2\n",
      "epoch 2623 completed in 27.33 seconds with avg training loss 305.82\n",
      "epoch 2624 completed in 27.13 seconds with avg training loss 305.85\n",
      "epoch 2625 completed in 27.08 seconds with avg training loss 306.14\n",
      "model not saved. validation loss was  804.68583984375\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 2626 completed in 26.83 seconds with avg training loss 306.33\n",
      "epoch 2627 completed in 26.78 seconds with avg training loss 306.34\n",
      "epoch 2628 completed in 27.09 seconds with avg training loss 305.83\n",
      "epoch 2629 completed in 27.19 seconds with avg training loss 305.88\n",
      "epoch 2630 completed in 27.02 seconds with avg training loss 306.94\n",
      "model not saved. validation loss was  800.83267578125\n",
      "epoch 2631 completed in 27.06 seconds with avg training loss 306.31\n",
      "epoch 2632 completed in 27.34 seconds with avg training loss 305.87\n",
      "epoch 2633 completed in 27.14 seconds with avg training loss 306.36\n",
      "epoch 2634 completed in 26.88 seconds with avg training loss 306.71\n",
      "epoch 2635 completed in 26.81 seconds with avg training loss 305.9\n",
      "model not saved. validation loss was  797.62923828125\n",
      "epoch 2636 completed in 26.78 seconds with avg training loss 306.54\n",
      "epoch 2637 completed in 27.27 seconds with avg training loss 305.71\n",
      "epoch 2638 completed in 26.63 seconds with avg training loss 305.61\n",
      "epoch 2639 completed in 26.92 seconds with avg training loss 305.94\n",
      "epoch 2640 completed in 26.97 seconds with avg training loss 306.25\n",
      "model not saved. validation loss was  804.6554296875\n",
      "epoch 2641 completed in 26.52 seconds with avg training loss 305.86\n",
      "epoch 2642 completed in 26.84 seconds with avg training loss 306.03\n",
      "epoch 2643 completed in 26.61 seconds with avg training loss 306.54\n",
      "epoch 2644 completed in 26.8 seconds with avg training loss 305.86\n",
      "epoch 2645 completed in 26.98 seconds with avg training loss 306.02\n",
      "model not saved. validation loss was  801.2431445312501\n",
      "epoch 2646 completed in 26.95 seconds with avg training loss 306.59\n",
      "epoch 2647 completed in 27.08 seconds with avg training loss 306.03\n",
      "epoch 2648 completed in 27.2 seconds with avg training loss 306.14\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-288-48042bf991f8>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     17\u001b[0m     \u001b[0mtrain_loss\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;36m0\u001b[0m \u001b[1;31m# mean loss per epoch\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     18\u001b[0m     \u001b[0mt0\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtime\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtime\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m# to estimate claculation time per epoch\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 19\u001b[1;33m     \u001b[1;32mfor\u001b[0m \u001b[0mi\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtargets\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;32min\u001b[0m \u001b[0menumerate\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtrain_loader\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m \u001b[1;31m# batch iterator\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     20\u001b[0m         \u001b[0minputs\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtargets\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0minputs\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mto\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdevice\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtargets\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mto\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdevice\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m# batch to gpu\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     21\u001b[0m         \u001b[0moptimizer\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mzero_grad\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m# zero gradients\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Miniconda3\\envs\\kevintorch-env\\lib\\site-packages\\torch\\utils\\data\\dataloader.py\u001b[0m in \u001b[0;36m__next__\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    312\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnum_workers\u001b[0m \u001b[1;33m==\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m:\u001b[0m  \u001b[1;31m# same-process loading\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    313\u001b[0m             \u001b[0mindices\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnext\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msample_iter\u001b[0m\u001b[1;33m)\u001b[0m  \u001b[1;31m# may raise StopIteration\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 314\u001b[1;33m             \u001b[0mbatch\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcollate_fn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mi\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mindices\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    315\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpin_memory\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    316\u001b[0m                 \u001b[0mbatch\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mpin_memory_batch\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mbatch\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Miniconda3\\envs\\kevintorch-env\\lib\\site-packages\\torch\\utils\\data\\dataloader.py\u001b[0m in \u001b[0;36m<listcomp>\u001b[1;34m(.0)\u001b[0m\n\u001b[0;32m    312\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnum_workers\u001b[0m \u001b[1;33m==\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m:\u001b[0m  \u001b[1;31m# same-process loading\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    313\u001b[0m             \u001b[0mindices\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnext\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msample_iter\u001b[0m\u001b[1;33m)\u001b[0m  \u001b[1;31m# may raise StopIteration\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 314\u001b[1;33m             \u001b[0mbatch\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcollate_fn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdataset\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mi\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mindices\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    315\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpin_memory\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    316\u001b[0m                 \u001b[0mbatch\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mpin_memory_batch\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mbatch\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m<ipython-input-113-7cee427a5b05>\u001b[0m in \u001b[0;36m__getitem__\u001b[1;34m(self, index)\u001b[0m\n\u001b[0;32m     16\u001b[0m         \u001b[1;32melif\u001b[0m \u001b[0mmode\u001b[0m\u001b[1;33m==\u001b[0m\u001b[1;34m'test'\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     17\u001b[0m             \u001b[0mfname1\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;34m'test-comp-%05d.npy'\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 18\u001b[1;33m         \u001b[0mimg\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mPIL\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mImage\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mopen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m+\u001b[0m\u001b[0mfname\u001b[0m\u001b[1;33m%\u001b[0m\u001b[0mindex\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     19\u001b[0m         \u001b[0mvect\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mload\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpath\u001b[0m\u001b[1;33m+\u001b[0m\u001b[0mfname1\u001b[0m\u001b[1;33m%\u001b[0m\u001b[0mindex\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     20\u001b[0m         transform = transforms.Compose([#transforms.Scale((227,227)),\n",
      "\u001b[1;32m~\\Miniconda3\\envs\\kevintorch-env\\lib\\site-packages\\PIL\\Image.py\u001b[0m in \u001b[0;36mopen\u001b[1;34m(fp, mode)\u001b[0m\n\u001b[0;32m   2578\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2579\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[0mfilename\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2580\u001b[1;33m         \u001b[0mfp\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mbuiltins\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mopen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mfilename\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m\"rb\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2581\u001b[0m         \u001b[0mexclusive_fp\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;32mTrue\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2582\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "criterion = nn.MSELoss()\n",
    "optimizer = optim.Adam(model.parameters(), lr=0.001)\n",
    "scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer,verbose=True)\n",
    "train_losses = []\n",
    "val_losses = []\n",
    "min_val_loss = 999999.9 # for early stopping \n",
    "\n",
    "# set up the training loop and dataset iterator \n",
    "k = 200 #size of batch \n",
    "N = 5000 #number epochs\n",
    "b = int(len(train_data)/k) #number of batches\n",
    "train_loader = DataLoader(train_data , batch_size = k, shuffle = True) #batch data loader\n",
    "test_loader = DataLoader(test_data, batch_size = k, shuffle = False)\n",
    "\n",
    "# train the network N epochs through the dataset \n",
    "for epoch in range(N): # epoch iterator \n",
    "    train_loss = 0 # mean loss per epoch \n",
    "    t0 = time.time() # to estimate claculation time per epoch \n",
    "    for i, (inputs, targets) in enumerate(train_loader): # batch iterator \n",
    "        inputs, targets = inputs.to(device), targets.to(device) # batch to gpu\n",
    "        optimizer.zero_grad() # zero gradients\n",
    "        outputs = model(inputs) # model prediction\n",
    "        loss = criterion(outputs, targets)  # loss computation\n",
    "        loss.backward() # backpropagation\n",
    "        optimizer.step() # gradient descent \n",
    "        train_loss += loss.cpu().data.item()/k # average loss across a batch \n",
    "    train_loss /= i # average loss across an epoch \n",
    "    print('epoch {} completed in {} seconds with avg training loss {}'.format(epoch,\n",
    "                                                                              round((time.time()-t0), 2),\n",
    "                                                                              round(train_loss, 2)\n",
    "                                                                             ))\n",
    "    train_losses.append(train_loss)\n",
    "    scheduler.step(train_loss) # possibly modify the learning rate if avg training loss has plateaued\n",
    "    \n",
    "    if epoch%5==0:  # early stopping \n",
    "        # calculate the avg validation loss per batch\n",
    "        val_loss = 0 \n",
    "        for i, (inputs, targets) in enumerate(test_loader):\n",
    "            inputs, targets = inputs.to(device), targets.to(device) # batch to gpu\n",
    "            outputs = model(inputs) # model prediction\n",
    "            loss = criterion(outputs, targets)\n",
    "            val_loss += loss.cpu().data.item()/k # average loss over validation batch \n",
    "        val_loss /= i \n",
    "        val_losses.append(val_loss)\n",
    "        if val_loss < min_val_loss:\n",
    "            min_val_loss = val_loss \n",
    "            torch.save(model.state_dict(),'earlystop.pt') # save partially trained model \n",
    "            print('model saved (early stop). ', 'validation loss was ', min_val_loss)\n",
    "        else:\n",
    "            print('model not saved. validation loss was ', val_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0,0.5,'loss')"
      ]
     },
     "execution_count": 237,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtcAAAIaCAYAAADr+GE3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3XuUnOdB5/nfY+tmSbZs6+b4Ksd3O7YVR4YEBsZMwCQ+mTiwuThLwCQZTEJgJjuHhckyu8yyZ8/JLMPMgb0wk1lMkt1MSCBkCGwCJDDAmXMIiWxLvju+xE7kiyTLV9mWrcuzf7zVVHV3tdSSn+pStz6fc+pU9fu+3XpUKlV/++3nfd9Saw0AAPDqHTfuAQAAwEIhrgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKCRReMewKuxZs2aumHDhnEPAwCABe6WW255sta69lDbzeu43rBhQzZv3jzuYQAAsMCVUh6ZzXamhQAAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANDIonEPAAAOR63Jtm3J5s3929atySmnJJdc0r9dfHF3W7ly3CMGjiXiGoCj2hNPTA7pzZuT7du7dYsWJZdfnlx3XfLMM8k99yR//MfJvn39zz/rrH5sD4b3unVJKeP5OwELl7gG4Kjx5JPJLbf0I/qb30wefbRbd9xxyaWXJm99a7JpU3e74orkhBMmf429e5MHHkjuvbeL7Xvu6R7ffHOye3d/u8E93YPhfc45yfHHz93fGVhYSq113GM4Yps2baqbN28e9zAAOALPPJPceuvkkH744f76iy7qR/SmTcnGja9uisfEdJKJ2B4M74k94UmybFly4YXTw/vCC7t1wLGplHJLrXXTIbcT1wCM2u7dyW23TQ7p++/vr3/tayeH9FVXJatWzd34nnqqH9yD4f3tb3dRnnRTSM49d/q87ksu6faCAwvbbOPatBAAmnrppe4Aw4mI3ry5C9WJSD3rrC6gf/qnu/s3vCFZvXqsQ86ppybf933dbdCePcm3vjV5L/c99yRf+1ry8sv97davnz6n+5JLkjPPNK8bjjXiGoAj9soryR139CN68+bkzjuT/fu79evXJ1dfnbz73f290uvXj3fMh2PZsm5e9xVXTF6+f3/yyCP96J4I7899Lnn66f52K1f2z1oyGN7nn58sXjy3fxdgbpgWAsCs7NuX3H335JC+/fYusJNu7/Pg1I6rr05OP/3Y2nNba7Jjx+SpJRPh/d3v9rdbtKgL7KkHU150UXLiieMbPzAz00IAOGL793fTIQZDesuWbspHkpx0UhfQH/1oP6TPOefYCulhSun2zK9fn/zDfzh53e7dXWRPDe+ppw4888zhZzFx6kCYH+y5BjjG1Zo8+ODkkL711v5p61as6A4wnIjoTZuS887rTo3Hq7d3b/f8DzuLydRTBw6b171hg1MHwlxwthAApqm1mys8eEGWW27pTouXdHOMN26cHNIXXSTexqHW7hzfUw+mvOeeyacOXLp05lMHTj0HOHDkxDUAefTR6Vc3fPLJbt3ixd2BeoMhfemlDrSbD55+evpe7olTBx440G0zcerAwb3dE/F96qnjHT/MR+ZcAxxjduyYHtKPP96tO/745LLLkre/vR/Sl1/e7fVk/jnllORNb+pugyZOHTg1vP/yL7t1E9atmz6n26kDoQ1xDTAPPfXU5MuEb96cfOc73bpSumj6kR/pn7njyiuT5cvHO2ZGbzanDhwM789/fvKpA1es6Af3YHifd16yZMnc/l1gvjItBOAo99xz0y8T/tBD/fUXXDD5FHivf73TuTE7tSY7dw6f1z311IHnnTd9XvfFF3utcewwLQRgHnrhhe6Ud4Mhfd99/fXnnNNN67jppv5lwl16myNVSjdFZN264acOvO++6dH9J38y+dSBZ5wx/JLw69ebYsKxSVwDjMmePd1FWAYvE3733f0D0s44owvo972vf5nwtWvHO2aOHStXdq+5N7xh8vK9e7vfnEy9SM7v/u7kUweefPL06SUXX9wdZOnsMyxkpoUAzIG9e7vLgg+G9B139PcArl3bP9BwIqRPP328Y4bDMXHqwGFXp3ziif52E6cOnHowpVMHcrRzKj6AMdm3rwuKwZDeujV5+eVu/SmnTJ4jvWlTctZZfoXOwjVx6sCpZzF56KHJpw7csGH43u7Vq8c6fEgirgHmxIEDyf33Tw7p225LXnyxW3/iid1e6ImIvvrq7tfiQhq6qVH33z99Xvd9900+deDatcMvCX/mma4UytxxQCNAY7V2F+kYvEz4Lbckzz/frT/hhO4Aw5/5mX5IX3CBb/4wk2XLuvOtX3755OX793enlpx6FpPf//3uNJQTVqzoriA6NbzPP9+pAxkfe64Bhqi1OxXZ1IuyTJwTeMmS/mXCJ0L64ou7U5YBo1Frd4XRqXO677mnf573pDtg8vzzp08vufji5KSTxjd+5jd7rgEOw+OPTw/pHTu6dYsWdXvW3vnOfkhfdpk9YzDXSummiKxdm/zgD05et3t3d3XKqeH95S93BxRPOOOM4ZeEP+0007VoQ1wDx5wnn5we0o8+2q077rjk0kuT667rn73jiiu6X18DR6+VK7tpWVddNXn54KkDB+d1f+pT/SldSbJq1fBLwjt1IIdLXAML2jPPTL9M+MMP99dfdFFyzTX9kN64sZvHCSwMixd3/88vumjy8lqTxx6bHt1/+qfJJz/Z327Jku40gVPD+6KLnDqQ4cy5BhaM55/vztQxGNL3399f/9rX9qd1TFwmfNWq8Y0XODo980w/uAfDe+qpA885Z/hZTJw6cGEy5xpY0F58sTt39OBlwu+9t9sblXTnjb766uT97+9flOXUU8c7ZmB+OPnk5I1v7G6D9uxJHnhg+sGUf/VXyUsv9bdbu3b6wZSXXNK9Lzl70MInroGj3ssvd1czHDyX9F13dafrSroDka6+Ornhhn5Ir18/3jEDC8+yZcnrXtfdBh04kDzyyPSL5HzhC8muXf3tli/vn7VkMLwvuMAB0guJaSHAUWXv3uTuuyeH9O2394/2X7168mXCN23qLhPuKH/gaLRz5/TpJcNOHXjeef3oftvbkn/wD8Y3ZoYzLQQ46u3f312JbTCkt2zpX5lt1aounv/5P++H9DnnCGlg/pjp1IEvvNC9/00N7698pZuWIq7nL3ENzIkDB5IHH5wc0rfe2n2DSbozdLzhDcnP/Vw/pM87z/xEYGFasWL4qQP37UteeWU8Y6INcQ00V2s3/3DqZcKffbZbv2xZd6aOD3ygH9IXXeRcsgCLFrnS63znnw94VSbOFTsY0ps39w/iWbw4ufLK5L3v7Yf0ZZf55gHAwuTbG3BYtm+ffnXDJ57o1h1/fHcU/Tve0Q/pyy9Pli4d75gBYK6Ia2BGu3ZNv7rhd7/brSulO6r92mv7F2a58kpXLAPg2CaugSTdfOhbb50c0g891F9/wQXd0esTIf361ycrV45vvABwNBLXcAx6+unu6oa33dbfM33fff31GzZ0EX3TTV1IX3VVd2ooAODgxDUsYBNn7diyZfLtkUf625xxRhfS73tfF9JveEOyZs34xgwA89nI4rqUcnOStyXZUWt93cDyX0jy80n2Jfn/aq2/1Fv+sSQfTLI/yT+ttf7ZqMYGC9Err3QXIJgI6Ntu6/ZOP/NMt76U7nR3b3pT8uEPJxs3dnOkTzttvOMGgIVklHuuP5nk/0jy6YkFpZQfSnJ9kitqrS+XUtb1ll+a5IYklyU5PcnXSikX1lr3j3B8MG8980wXzoN7o++6q3+J8OXLkyuuSG64oYvojRu7s3isWDHecQPAQjeyuK61/k0pZcOUxR9O8vFa68u9bXb0ll+f5Pd6y79dSnkgyfck+dtRjQ/mg1q7s3NM7ImeCOmHH+5vs25dd3Dhj/5od79xY3L++S7IAgDjMNdzri9M8gOllP81yZ4kv1hr/WaSM5J8fWC7bb1lcMzYu3fytI6J29NPd+tLSS68MPne701+9mf7e6RN6wCAo8dcx/WiJKckeWOSq5N8vpTy2iRlyLZ12BcopdyU5KYkOfvss0c0TBitZ58dPq3jlVe69Sec0E3rePe7+xF9+eWmdQDA0W6u43pbkj+stdYk3yilHEiyprf8rIHtzkzy2LAvUGv9RJJPJMmmTZuGBjgcLWpNtm2bfJDhli3Jt7/d32bt2m46x0c/2kX061/fnVPatA4AmH/mOq7/c5J/lOSvSikXJlmS5MkkX0ryn0op/zbdAY0XJPnGHI8NXpW9e5N7750+reOpp7r1pXTRfPXVyc/8zORpHWXY724AgHlnlKfi+2ySa5KsKaVsS/KrSW5OcnMp5c4kryS5sbcX+65SyueT3J3uFH0fcaYQjmbPPTd9Wsedd/andSxb1k3reOc7J0/rcEVDAFjYSte289OmTZvq5s2bxz0MFrBak0cfnb43+sEH+9usWdM/S8fE7cILk0Uu0QQAC0Yp5ZZa66ZDbefbP/Ts2zd8WseuXf1tLriguxT4Bz7QD+nXvMa0DgCgI645Jj3/fHL77ZMPNLzzzuTll7v1S5d20zh+7Mf6Bxlefnly4onjHTcAcHQT1yxotSaPPTZ9b/QDD/S3Wb26i+df+IX+3uiLLjKtAwA4fPKBBWPfvuS++6aH9JNP9rc5//wunn/6p/shffrppnUAAG2Ia+al3bsnT+vYsiW5445kz55u/dKlyetel1x/fT+ir7giOemk8Y4bAFjYxDVHtVqTxx8fPq1j4kQ3p57aTev4yEcmT+tYvHi8YwcAjj3imqPG/v3Jt77Vv4rhxG3nzv42r31tF88/9VP9kD7jDNM6AICjg7hmLHbv7qZxTJ3W8dJL3folS7ppHf/4H0+e1rFq1XjHDQBwMOKakRs2reP++/vTOk45pZvW8eEP90P64otN6wAA5h9xTTP793fRPDWkt2/vb3PuuV08/8RP9K9qeOaZpnUAAAuDuOaIvPDC9Gkdt9/en9axeHE3reO66yZP6zj55PGOGwBglMQ1h7R9+/SDDL/1rf60jpNP7vZCf+hDk6d1LFky3nEDAMw1cc3f27+/O8Xd1GkdTzzR32bDhi6e3/vefkiffbZpHQAAibg+Zr344vBpHS++2K1fvDi57LLkLW/pR/SVV5rWAQBwMOL6GLBjRz+gJ6Z3fOtbyYED3fpVq7p4vummfkhfcolpHQAAh0tcLyAHDgyf1vH44/1tzjmni+f3vKcf0uecY1oHAEAL4nqeevHF5M47p0/reOGFbv2iRd20jmuvnTyt45RTxjtuAICFTFzPAzt3To7o225L7ruvP63jpJO6eP7gB/shfemlydKl4x03AMCxRlwfRQ4cSB58cPq0jsce629z9tldPL/rXf2Q3rDBtA4AgKOBuB6Tl15K7rpr8kGGt9+e7N7drV+0qDuo8Id/ePK0jlNPHe+4AQCYmbieA08+OX1v9L33dueVTpITT+zi+f3vnzytY9my8Y4bAIDDI64bOnAgeeih6SH96KP9bc46q4vnH//xydM6jjtubMMGAKARcX2E9uzpT+uYuG3dmjz/fLf++OO7aR0/9EPdpcEnpnWsXj3ecQMAMDri+jB9+9vJ29+e3HPP5GkdV16Z3Hhjf2/0ZZeZ1gEAcKwR14fptNO6aRzveEc/pM8917QOAADE9WE74YTkj/943KMAAOBoZH8rAAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0MjI4rqUcnMpZUcp5c4h636xlFJLKWt6H5dSym+VUh4opdxeSrlqVOMCAIBRGeWe608mecvUhaWUs5L8SJLvDCx+a5ILerebkvz2CMcFAAAjMbK4rrX+TZKnhqz6d0l+KUkdWHZ9kk/XzteTnFxKec2oxgYAAKMwp3OuSylvT/JorXXrlFVnJPnuwMfbessAAGDeWDRXf1ApZXmSX0ly7bDVQ5bVIctSSrkp3dSRnH322c3GBwAAr9Zc7rk+L8m5SbaWUh5OcmaSW0spp6XbU33WwLZnJnls2BeptX6i1rqp1rpp7dq1Ix4yAADM3pzFda31jlrrulrrhlrrhnRBfVWt9YkkX0ryU72zhrwxybO11sfnamwAANDCKE/F99kkf5vkolLKtlLKBw+y+ZeTPJTkgST/McnPjWpcAAAwKiObc11rfe8h1m8YeFyTfGRUYwEAgLngCo0AANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAjs4rrUso/K6WcVDq/U0q5tZRy7agHBwAA88ls91x/oNb6XJJrk6xN8v4kHx/ZqAAAYB6abVyX3v11SX631rp1YBkAAJDZx/UtpZQ/TxfXf1ZKOTHJgdENCwAA5p9Fs9zug0k2Jnmo1vpiKeXUdFNDAACAntnuuX5Tkvtqrc+UUt6X5F8meXZ0wwIAgPlntnH920leLKVcmeSXkjyS5NMjGxUAAMxDs43rfbXWmuT6JL9Za/3NJCeOblgAADD/zHbO9fOllI8l+ckkP1BKOT7J4tENCwAA5p/Z7rl+T5KX053v+okkZyT59ZGNCgAA5qFZxXUvqD+TZFUp5W1J9tRazbkGAIABs738+buTfCPJu5K8O8nflVLeOcqBAQDAfDPbOde/kuTqWuuOJCmlrE3ytSR/MKqBAQDAfDPbOdfHTYR1z67D+FwAADgmzHbP9Z+WUv4syWd7H78nyZdHMyQAAJifZhXXtdb/vpTy3yT5/iQlySdqrV8c6cgAAGCeme2e69Rav5DkCyMcCwAAzGsHjetSyvNJ6rBVSWqt9aSRjAoAAOahg8Z1rdUlzgEAYJac8QMAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhkZHFdSrm5lLKjlHLnwLJfL6XcW0q5vZTyxVLKyQPrPlZKeaCUcl8p5UdHNS4AABiVUe65/mSSt0xZ9tUkr6u1XpHkW0k+liSllEuT3JDkst7n/F+llONHODYAAGhuZHFda/2bJE9NWfbntdZ9vQ+/nuTM3uPrk/xerfXlWuu3kzyQ5HtGNTYAABiFcc65/kCSr/Qen5HkuwPrtvWWTVNKuamUsrmUsnnnzp0jHiIAAMzeWOK6lPIrSfYl+czEoiGb1WGfW2v9RK11U61109q1a0c1RAAAOGyL5voPLKXcmORtSd5ca50I6G1JzhrY7Mwkj8312AAA4NWY0z3XpZS3JPnlJG+vtb44sOpLSW4opSwtpZyb5IIk35jLsQEAwKs1sj3XpZTPJrkmyZpSyrYkv5ru7CBLk3y1lJIkX6+1fqjWelcp5fNJ7k43XeQjtdb9oxobAACMQunPzJh/Nm3aVDdv3jzuYQAAsMCVUm6ptW461Hau0AgAAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKCRkcV1KeXmUsqOUsqdA8tOLaV8tZRyf+/+lN7yUkr5rVLKA6WU20spV41qXAAAMCqj3HP9ySRvmbLsXyT5i1rrBUn+ovdxkrw1yQW9201JfnuE4wIAgJEYWVzXWv8myVNTFl+f5FO9x59K8o6B5Z+una8nObmU8ppRjQ0AAEZhrudcr6+1Pp4kvft1veVnJPnuwHbbessAAGDeOFoOaCxDltWhG5ZyUyllcyll886dO0c8LAAAmL25juvtE9M9evc7esu3JTlrYLszkzw27AvUWj9Ra91Ua920du3akQ4WAAAOx1zH9ZeS3Nh7fGOSPxpY/lO9s4a8McmzE9NHAABgvlg0qi9cSvlskmuSrCmlbEvyq0k+nuTzpZQPJvlOknf1Nv9ykuuSPJDkxSTvH9W4AABgVEYW17XW986w6s1Dtq1JPjKqsQAAwFw4Wg5oBACAeU9cAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCPiGgAAGhHXAADQiLgGAIBGxDUAADQirgEAoBFxDQAAjYhrAABoRFwDAEAj4hoAABoR1wAA0Ii4BgCARsQ1AAA0Iq4BAKARcX0kdu4c9wgAADgKLRr3AOadJ55IXvOa7nbllcnGjd39lVcmF16YHH/8uEcIAMCYiOvDtWRJ8hu/kWzdmmzZknzta8m+fd26E05IXve6ycF9xRXJSSeNd8wAAMyJUmsd9xiO2KZNm+rmzZvHO4hXXknuuacL7Yng3ro1eeqp/javfW0/uCfuzz47KWV84waObvv3J/ffP/l95ZlnknXrkvXr+7fBj9etS04+2XsLwAiUUm6ptW461Hb2XL9aS5b091JPqDV59NHpwf3FL3brku4b4MTnTQT3ZZclS5eO5+/BwvbKK8kLLyS7d3f3E49ffLF7Da9Y0d2WL5/82DSnufHcc8ntt3fvExPvGXfembz0Urd+0aLk0kuTtWuThx5K/vZvkyefTA4cmP61lizpB/fU8J4a5WvW+DcGaMye67m0e3dyxx2Tg/v227vASbpvoBdfPH0u97p14x13S7V202j27ev2zE08ns3HR/I5+/d3f+bixd1tyZL+49ncZtp+0aL2ewf37+9eC1MDeOLxsDg+1OOJjyemLh2uZcsmB/ewCJ9p2aG2OeGEY28Pa63JI49MjuitW7tgnnDqqZP//2/cmFxySfdaHLR/f7JrV7J9e3fbsaP/eNjHe/dOH89xx3WBfagIn7j3wz9wDJvtnmtxPW779ycPPjj5G+3Wrcm2bf1tph48uXFjcsEFr36PU63dnrEXXuiibvD+1Szbs2fmKB62p22+WrRo9jE+eNu7d3gQ79lzeH/+smX9UF258sgeL1/e36s9+G86eJvt8om9rLNVSvfnDwb3ypXJaaclp5/e3V7zmsmP16zpgnA+2LMnueuuyf+vt25Nnn22W19K9/946m+wzjij/Q8dtXZ/7sHie/DjF14Y/nVWrZoc3mvXdrc1a6bfr1mzcGN84gebHTu6s0dN3A8+fvrp5MQTux+Wpt5Wr5788cqVC+MHzT17uimRu3Z191MfD368f//0H+ROO23yx8uXj/tvBJOI6/lu167pe7fuvru/9+mEE5LLL+++GV8J6FjzAAANXUlEQVR8cfdGNZsAnrr+cC1dOn0v5NT7Zcv6e3cXLep+CJh4POzj2WxzJJ8zsSzpnreJ2yuvTP74ULfD3f5Qn7N48eTQPZI4XrHi6Pt1/oED/R/WjjTQn3++OyPPY49NPm5hwqJF/bP1TET3sAhfvXpuI3z79skRvWVLct993f/LpPv3uuKKyRF9+eXd8qPRCy9Mju2ZQnznzu7faabvIyeeePAAn3q/atV4InP//u7vMSyShwX0rl3D/86ldK+9tWuTU07pfmieiMqDvd8uWjQ8wg91W7VqNK/zw4nkwccH+wF78eLuuZn4weK44/qvp6efHv45K1dOju2ZInz9+m7bo93+/d1r4tlnu6lgU+8HHyeTv5cNfk+dzcdH8jkzfbwQfvBrRFwvRBMHTw4G95YtkyPkhBOGz5091LLZbG8OLnNpz55+aD/2WPL448MfD/vGvHjx7CP8cL5x7NvXRfPU3zRt397f5qyzpk/tOu+8+bPH/XDt29f9G+zc2c0DnwjQicdT73fuTF5+efjXWrSoH9uzCfLVq6dPl0m6H/SeeurQkTzxeNeumX+rNhHL69b1xzXxeOr96tUzv0fu2dM9T4OBOixap96ef37m576ULuJns3f8hBO6A2JnE82HE8nD/rxh65Yvn/n/2iuvTP8BbvD2xBP9x7t2Df8ay5fPLsLXr+9+6Duc//e1dj8cHSyGD3X/3HMH/7ecUEp/fFN/+zsuxx3X/T8b9W3p0tlve8453WtxjonrY0Wt3RvmkiXdm+dC/QYOM3nppX6EzxTgjz3W/T+ZasmSg0f44sX94yS2bu0OMpwIwyVLuoMMByP6yiu7mGBmtXZ7xmeK72H3w36LMWHVqi62V6/uvu7E58wUy6eeeuhIHozlRWM+7n/v3uFRfqjbM8/M/BuFpB/Jswnj2UbyXNi7t/s3ninEB4P8ySeHPwfLlk2P8OOOmzmcn3tudnG7YkV36t1VqybfD1s207oVK4Z/Hz9wYPoxRXv3HvzjVtsM/kb2cG4vvzzzulf7w8Ijj3RnXZtj4hpg0Esv9YN7pgh//PHhEb5mzfSDDC++eCx7To5J+/b190TPtHd8164uTA62l3n16mPn32z//v6e6olpKYN7uVesWPi/7t+3r3ttHCrEt2/v4vVIg/ikk7q9zeP+QWw+2b//yIJ94vbjPz6WOfniGuBIvPhiF9mPP94F+WWXdXuyF3qIAHBQznMNcCSWL+/mSJ933rhHAsA8ZIIuAAA0Iq4BAKARcQ0AAI2IawAAaERcAwBAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGhEXAMAQCNjietSyn9XSrmrlHJnKeWzpZRlpZRzSyl/V0q5v5TyuVLKknGMDQAAjtScx3Up5Ywk/zTJplrr65Icn+SGJP86yb+rtV6Q5OkkH5zrsQEAwKsxrmkhi5KcUEpZlGR5kseT/KMkf9Bb/6kk7xjT2AAA4IjMeVzXWh9N8m+SfCddVD+b5JYkz9Ra9/U225bkjLkeGwAAvBrjmBZySpLrk5yb5PQkK5K8dcimdYbPv6mUsrmUsnnnzp2jGygAABymcUwL+eEk36617qy17k3yh0m+L8nJvWkiSXJmkseGfXKt9RO11k211k1r166dmxEDAMAsjCOuv5PkjaWU5aWUkuTNSe5O8l+SvLO3zY1J/mgMYwMAgCNWah06+2K0f2gp/3OS9yTZl+S2JP8k3Rzr30tyam/Z+2qtLx/i6+xM8shoRzujNUmeHNOffazwHM8Nz/PoeY7nhud59DzHc8PzPHpH8hyfU2s95LSJscT1QlBK2Vxr3TTucSxknuO54XkePc/x3PA8j57neG54nkdvlM+xKzQCAEAj4hoAABoR10fuE+MewDHAczw3PM+j5zmeG57n0fMczw3P8+iN7Dk25xoAABqx5xoAABoR1wdRSnlLKeW+UsoDpZR/MWT90lLK53rr/66UsmHuRzm/lVLOKqX8l1LKPaWUu0op/2zINteUUp4tpWzp3f6ncYx1viulPFxKuaP3HG4esr6UUn6r93q+vZRy1TjGOV+VUi4aeI1uKaU8V0r56JRtvJaPQCnl5lLKjlLKnQPLTi2lfLWUcn/v/pQZPvfG3jb3l1JunLtRzy8zPMe/Xkq5t/d+8MVSyskzfO5B31vom+F5/lellEcH3heum+FzD9okdGZ4jj838Pw+XErZMsPnNnktmxYyg1LK8Um+leRHkmxL8s0k76213j2wzc8luaLW+qFSyg1JfqzW+p6xDHieKqW8Jslraq23llJOTHJLkndMeZ6vSfKLtda3jWmYC0Ip5eEkm2qtQ8/r2XtD/4Uk1yX53iS/WWv93rkb4cLRe/94NMn31lofGVh+TbyWD1sp5QeT7E7y6Vrr63rL/rckT9VaP94LjVNqrb885fNOTbI5yaYkNd37yxtqrU/P6V9gHpjhOb42yV/WWveVUv51kkx9jnvbPZyDvLfQN8Pz/K+S7K61/puDfN4hm4TOsOd4yvrfSPJsrfXXhqx7OA1ey/Zcz+x7kjxQa32o1vpKugvcXD9lm+uTfKr3+A+SvLl31Ulmqdb6eK311t7j55Pck+6CQsy969O9GdVa69eTnNz74YfD9+YkDw6GNUeu1vo3SZ6asnjw/fdTSd4x5FN/NMlXa61P9YL6q0neMrKBzmPDnuNa65/XWvf1Pvx6kjPnfGALzAyv5dmYTZOQgz/HvUZ7d5LPjnIM4npmZyT57sDH2zI9+v5+m94b0LNJVs/J6Bag3rSa1yf5uyGr31RK2VpK+Uop5bI5HdjCUZP8eSnlllLKTUPWz+Y1z+zckJnfvL2W21hfa3086X5IT7JuyDZe0+18IMlXZlh3qPcWDu3ne9Nvbp5hipPXchs/kGR7rfX+GdY3eS2L65kN2wM9dQ7NbLZhFkopK5N8IclHa63PTVl9a7pLjl6Z5H9P8p/nenwLxPfXWq9K8tYkH+n96myQ13MDpZQlSd6e5PeHrPZanlte0w2UUn4lyb4kn5lhk0O9t3Bwv53kvCQbkzye5DeGbOO13MZ7c/C91k1ey+J6ZtuSnDXw8ZlJHptpm1LKoiSrcmS/7jmmlVIWpwvrz9Ra/3Dq+lrrc7XW3b3HX06yuJSyZo6HOe/VWh/r3e9I8sV0v2YcNJvXPIf21iS31lq3T13htdzU9olpS737HUO28Zp+lXoHgb4tyU/UGQ7SmsV7CwdRa91ea91faz2Q5D9m+PPntfwq9Trtx5N8bqZtWr2WxfXMvpnkglLKub09UTck+dKUbb6UZOLo83emO/DDT5KHoTf/6XeS3FNr/bczbHPaxFz2Usr3pHvd7pq7Uc5/pZQVvQNGU0pZkeTaJHdO2exLSX6qdN6Y7oCPx+d4qAvBjHtGvJabGnz/vTHJHw3Z5s+SXFtKOaX3q/Zre8uYhVLKW5L8cpK311pfnGGb2by3cBBTjm35sQx//mbTJBzcDye5t9a6bdjKlq/lRUc8xAWud3T0z6d7Iz4+yc211rtKKb+WZHOt9UvpovD/KaU8kG6P9Q3jG/G89f1JfjLJHQOnxvkfkpydJLXWf5/uB5cPl1L2JXkpyQ1+iDls65N8sdd1i5L8p1rrn5ZSPpT8/fP85XRnCnkgyYtJ3j+msc5bpZTl6Y7m/9mBZYPPsdfyESilfDbJNUnWlFK2JfnVJB9P8vlSygeTfCfJu3rbbkryoVrrP6m1PlVK+V/ShUmS/Fqt1W8Xh5jhOf5YkqVJvtp77/h67+xYpyf5v2ut12WG95Yx/BXmhRme52tKKRvTTfN4OL33j8HneaYmGcNf4ag37Dmutf5OhhwLM6rXslPxAQBAI6aFAABAI+IaAAAaEdcAANCIuAYAgEbENQAANCKuAZiklHJNKeVPxj0OgPlIXAMAQCPiGmCeKqW8r5TyjVLKllLKfyilHF9K2V1K+Y1Syq2llL8opaztbbuxlPL1UsrtpZQv9q5YmFLK+aWUr5VStvY+57zel19ZSvmDUsq9pZTPTFxZEoCDE9cA81Ap5ZIk70ny/bXWjUn2J/mJJCuS3FprvSrJX6e7AlySfDrJL9dar0hyx8DyzyT5P2utVyb5viQTl7x/fZKPJrk0yWvTXU0VgENw+XOA+enNSd6Q5Ju9nconJNmR5ECSz/W2+X+T/GEpZVWSk2utf91b/qkkv19KOTHJGbXWLyZJrXVPkvS+3jdqrdt6H29JsiHJfx39XwtgfhPXAPNTSfKpWuvHJi0s5X+csl09xNeYycsDj/fH9wuAWTEtBGB++osk7yylrEuSUsqppZRz0r2vv7O3zX+b5L/WWp9N8nQp5Qd6y38yyV/XWp9Lsq2U8o7e11haSlk+p38LgAXGngiAeajWencp5V8m+fNSynFJ9ib5SJIXklxWSrklybPp5mUnyY1J/n0vnh9K8v7e8p9M8h9KKb/W+xrvmsO/BsCCU2o92G8MAZhPSim7a60rxz0OgGOVaSEAANCIPdcAANCIPdcAANCIuAYAgEbENQAANCKuAQCgEXENAACNiGsAAGjk/wcgPiM7XSI01AAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 864x648 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(range(0,5*len(val_losses),5),val_losses,color='blue')\n",
    "plt.plot(range(len(train_losses)),train_losses,color='red')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# diagnose output by decompression / visualization "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.state_dict() = torch.load('trained.pt')\n",
    "model.eval()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_loader = DataLoader(test_data , batch_size = k, shuffle = True) #batch data loader\n",
    "for i,(x,y) in enumerate(test_loader):\n",
    "    break\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_ = model(x.to(device)).cpu().detach().numpy()\n",
    "y = y.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "for n in range(3):\n",
    "    plt.plot(y[n],'x')\n",
    "    plt.plot(y_[n],'o')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "S = np.load('./data/sensing_matrix.npy')\n",
    "shape = x[0].numpy().shape[1:]\n",
    "_,M = shape\n",
    "N = 100\n",
    "L = 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Y = y_[0]\n",
    "Y = Y.reshape(L,N)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "pts = comp.unproject(F,shape,5,M)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "im = x[0].numpy().squeeze()\n",
    "for p in pts:\n",
    "    im = comp.draw_plus(p,im)\n",
    "plt.imshow(im,cmap='gray')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "mask = (pts.T[0]>0)&(pts.T[0]<150)&(pts.T[1]>0)&(pts.T[1]<150)\n",
    "np.where(mask)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}