L09_mlp/code/mlp-pytorch.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "STAT 479: Deep Learning (Spring 2019)  \n",
    "Instructor: Sebastian Raschka (sraschka@wisc.edu)  \n",
    "Course website: http://pages.stat.wisc.edu/~sraschka/teaching/stat479-ss2019/\n",
    "GitHub repository: https://github.com/rasbt/stat479-deep-learning-ss19"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sebastian Raschka \n",
      "\n",
      "CPython 3.6.8\n",
      "IPython 7.2.0\n",
      "\n",
      "torch 1.0.0\n"
     ]
    }
   ],
   "source": [
    "%load_ext watermark\n",
    "%watermark -a 'Sebastian Raschka' -v -p torch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MLP With Different Loss Functions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import torch\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "import numpy as np\n",
    "from torchvision import datasets\n",
    "from torchvision import transforms\n",
    "from torch.utils.data import DataLoader\n",
    "import torch.nn.functional as F\n",
    "import torch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Settings and Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image batch dimensions: torch.Size([100, 1, 28, 28])\n",
      "Image label dimensions: torch.Size([100])\n"
     ]
    }
   ],
   "source": [
    "##########################\n",
    "### SETTINGS\n",
    "##########################\n",
    "\n",
    "RANDOM_SEED = 1\n",
    "BATCH_SIZE = 100\n",
    "NUM_EPOCHS = 100\n",
    "DEVICE = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')\n",
    "\n",
    "\n",
    "##########################\n",
    "### MNIST DATASET\n",
    "##########################\n",
    "\n",
    "# Note transforms.ToTensor() scales input images\n",
    "# to 0-1 range\n",
    "train_dataset = datasets.MNIST(root='data', \n",
    "                               train=True, \n",
    "                               transform=transforms.ToTensor(),\n",
    "                               download=True)\n",
    "\n",
    "test_dataset = datasets.MNIST(root='data', \n",
    "                              train=False, \n",
    "                              transform=transforms.ToTensor())\n",
    "\n",
    "\n",
    "train_loader = DataLoader(dataset=train_dataset, \n",
    "                          batch_size=BATCH_SIZE, \n",
    "                          shuffle=True)\n",
    "\n",
    "test_loader = DataLoader(dataset=test_dataset, \n",
    "                         batch_size=BATCH_SIZE, \n",
    "                         shuffle=False)\n",
    "\n",
    "# Checking the dataset\n",
    "for images, labels in train_loader:  \n",
    "    print('Image batch dimensions:', images.shape)\n",
    "    print('Image label dimensions:', labels.shape)\n",
    "    break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def to_onehot(y, num_classes):\n",
    "    y_onehot = torch.FloatTensor(y.size(0), num_classes)\n",
    "    y_onehot.zero_()\n",
    "    tmp = y.view(-1, 1).long().to(torch.device('cpu'))\n",
    "    y_onehot.scatter_(1, tmp, 1).float()\n",
    "    return y_onehot"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Sigmoid Activation and MSE Loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 001/100 | Batch 000/600 | Cost: 0.2454\n",
      "Epoch: 001/100 | Batch 050/600 | Cost: 0.1050\n",
      "Epoch: 001/100 | Batch 100/600 | Cost: 0.0943\n",
      "Epoch: 001/100 | Batch 150/600 | Cost: 0.0919\n",
      "Epoch: 001/100 | Batch 200/600 | Cost: 0.0902\n",
      "Epoch: 001/100 | Batch 250/600 | Cost: 0.0900\n",
      "Epoch: 001/100 | Batch 300/600 | Cost: 0.0896\n",
      "Epoch: 001/100 | Batch 350/600 | Cost: 0.0888\n",
      "Epoch: 001/100 | Batch 400/600 | Cost: 0.0891\n",
      "Epoch: 001/100 | Batch 450/600 | Cost: 0.0890\n",
      "Epoch: 001/100 | Batch 500/600 | Cost: 0.0884\n",
      "Epoch: 001/100 | Batch 550/600 | Cost: 0.0884\n",
      "Epoch: 001/100 Train Cost: 0.0881\n",
      "Time elapsed: 0.23 min\n",
      "Epoch: 002/100 | Batch 000/600 | Cost: 0.0881\n",
      "Epoch: 002/100 | Batch 050/600 | Cost: 0.0881\n",
      "Epoch: 002/100 | Batch 100/600 | Cost: 0.0878\n",
      "Epoch: 002/100 | Batch 150/600 | Cost: 0.0872\n",
      "Epoch: 002/100 | Batch 200/600 | Cost: 0.0869\n",
      "Epoch: 002/100 | Batch 250/600 | Cost: 0.0874\n",
      "Epoch: 002/100 | Batch 300/600 | Cost: 0.0864\n",
      "Epoch: 002/100 | Batch 350/600 | Cost: 0.0857\n",
      "Epoch: 002/100 | Batch 400/600 | Cost: 0.0859\n",
      "Epoch: 002/100 | Batch 450/600 | Cost: 0.0860\n",
      "Epoch: 002/100 | Batch 500/600 | Cost: 0.0861\n",
      "Epoch: 002/100 | Batch 550/600 | Cost: 0.0848\n",
      "Epoch: 002/100 Train Cost: 0.0852\n",
      "Time elapsed: 0.45 min\n",
      "Epoch: 003/100 | Batch 000/600 | Cost: 0.0849\n",
      "Epoch: 003/100 | Batch 050/600 | Cost: 0.0854\n",
      "Epoch: 003/100 | Batch 100/600 | Cost: 0.0840\n",
      "Epoch: 003/100 | Batch 150/600 | Cost: 0.0848\n",
      "Epoch: 003/100 | Batch 200/600 | Cost: 0.0841\n",
      "Epoch: 003/100 | Batch 250/600 | Cost: 0.0839\n",
      "Epoch: 003/100 | Batch 300/600 | Cost: 0.0851\n",
      "Epoch: 003/100 | Batch 350/600 | Cost: 0.0826\n",
      "Epoch: 003/100 | Batch 400/600 | Cost: 0.0839\n",
      "Epoch: 003/100 | Batch 450/600 | Cost: 0.0821\n",
      "Epoch: 003/100 | Batch 500/600 | Cost: 0.0842\n",
      "Epoch: 003/100 | Batch 550/600 | Cost: 0.0836\n",
      "Epoch: 003/100 Train Cost: 0.0812\n",
      "Time elapsed: 0.68 min\n",
      "Epoch: 004/100 | Batch 000/600 | Cost: 0.0799\n",
      "Epoch: 004/100 | Batch 050/600 | Cost: 0.0805\n",
      "Epoch: 004/100 | Batch 100/600 | Cost: 0.0832\n",
      "Epoch: 004/100 | Batch 150/600 | Cost: 0.0815\n",
      "Epoch: 004/100 | Batch 200/600 | Cost: 0.0806\n",
      "Epoch: 004/100 | Batch 250/600 | Cost: 0.0796\n",
      "Epoch: 004/100 | Batch 300/600 | Cost: 0.0781\n",
      "Epoch: 004/100 | Batch 350/600 | Cost: 0.0790\n",
      "Epoch: 004/100 | Batch 400/600 | Cost: 0.0769\n",
      "Epoch: 004/100 | Batch 450/600 | Cost: 0.0778\n",
      "Epoch: 004/100 | Batch 500/600 | Cost: 0.0773\n",
      "Epoch: 004/100 | Batch 550/600 | Cost: 0.0776\n",
      "Epoch: 004/100 Train Cost: 0.0763\n",
      "Time elapsed: 0.91 min\n",
      "Epoch: 005/100 | Batch 000/600 | Cost: 0.0763\n",
      "Epoch: 005/100 | Batch 050/600 | Cost: 0.0735\n",
      "Epoch: 005/100 | Batch 100/600 | Cost: 0.0765\n",
      "Epoch: 005/100 | Batch 150/600 | Cost: 0.0756\n",
      "Epoch: 005/100 | Batch 200/600 | Cost: 0.0739\n",
      "Epoch: 005/100 | Batch 250/600 | Cost: 0.0728\n",
      "Epoch: 005/100 | Batch 300/600 | Cost: 0.0737\n",
      "Epoch: 005/100 | Batch 350/600 | Cost: 0.0733\n",
      "Epoch: 005/100 | Batch 400/600 | Cost: 0.0738\n",
      "Epoch: 005/100 | Batch 450/600 | Cost: 0.0713\n",
      "Epoch: 005/100 | Batch 500/600 | Cost: 0.0710\n",
      "Epoch: 005/100 | Batch 550/600 | Cost: 0.0703\n",
      "Epoch: 005/100 Train Cost: 0.0709\n",
      "Time elapsed: 1.14 min\n",
      "Epoch: 006/100 | Batch 000/600 | Cost: 0.0685\n",
      "Epoch: 006/100 | Batch 050/600 | Cost: 0.0713\n",
      "Epoch: 006/100 | Batch 100/600 | Cost: 0.0663\n",
      "Epoch: 006/100 | Batch 150/600 | Cost: 0.0703\n",
      "Epoch: 006/100 | Batch 200/600 | Cost: 0.0731\n",
      "Epoch: 006/100 | Batch 250/600 | Cost: 0.0688\n",
      "Epoch: 006/100 | Batch 300/600 | Cost: 0.0705\n",
      "Epoch: 006/100 | Batch 350/600 | Cost: 0.0688\n",
      "Epoch: 006/100 | Batch 400/600 | Cost: 0.0686\n",
      "Epoch: 006/100 | Batch 450/600 | Cost: 0.0686\n",
      "Epoch: 006/100 | Batch 500/600 | Cost: 0.0660\n",
      "Epoch: 006/100 | Batch 550/600 | Cost: 0.0680\n",
      "Epoch: 006/100 Train Cost: 0.0657\n",
      "Time elapsed: 1.37 min\n",
      "Epoch: 007/100 | Batch 000/600 | Cost: 0.0654\n",
      "Epoch: 007/100 | Batch 050/600 | Cost: 0.0639\n",
      "Epoch: 007/100 | Batch 100/600 | Cost: 0.0666\n",
      "Epoch: 007/100 | Batch 150/600 | Cost: 0.0642\n",
      "Epoch: 007/100 | Batch 200/600 | Cost: 0.0665\n",
      "Epoch: 007/100 | Batch 250/600 | Cost: 0.0641\n",
      "Epoch: 007/100 | Batch 300/600 | Cost: 0.0627\n",
      "Epoch: 007/100 | Batch 350/600 | Cost: 0.0637\n",
      "Epoch: 007/100 | Batch 400/600 | Cost: 0.0608\n",
      "Epoch: 007/100 | Batch 450/600 | Cost: 0.0594\n",
      "Epoch: 007/100 | Batch 500/600 | Cost: 0.0594\n",
      "Epoch: 007/100 | Batch 550/600 | Cost: 0.0612\n",
      "Epoch: 007/100 Train Cost: 0.0610\n",
      "Time elapsed: 1.60 min\n",
      "Epoch: 008/100 | Batch 000/600 | Cost: 0.0614\n",
      "Epoch: 008/100 | Batch 050/600 | Cost: 0.0604\n",
      "Epoch: 008/100 | Batch 100/600 | Cost: 0.0611\n",
      "Epoch: 008/100 | Batch 150/600 | Cost: 0.0575\n",
      "Epoch: 008/100 | Batch 200/600 | Cost: 0.0592\n",
      "Epoch: 008/100 | Batch 250/600 | Cost: 0.0644\n",
      "Epoch: 008/100 | Batch 300/600 | Cost: 0.0572\n",
      "Epoch: 008/100 | Batch 350/600 | Cost: 0.0594\n",
      "Epoch: 008/100 | Batch 400/600 | Cost: 0.0561\n",
      "Epoch: 008/100 | Batch 450/600 | Cost: 0.0615\n",
      "Epoch: 008/100 | Batch 500/600 | Cost: 0.0575\n",
      "Epoch: 008/100 | Batch 550/600 | Cost: 0.0576\n",
      "Epoch: 008/100 Train Cost: 0.0568\n",
      "Time elapsed: 1.82 min\n",
      "Epoch: 009/100 | Batch 000/600 | Cost: 0.0561\n",
      "Epoch: 009/100 | Batch 050/600 | Cost: 0.0523\n",
      "Epoch: 009/100 | Batch 100/600 | Cost: 0.0550\n",
      "Epoch: 009/100 | Batch 150/600 | Cost: 0.0548\n",
      "Epoch: 009/100 | Batch 200/600 | Cost: 0.0556\n",
      "Epoch: 009/100 | Batch 250/600 | Cost: 0.0556\n",
      "Epoch: 009/100 | Batch 300/600 | Cost: 0.0530\n",
      "Epoch: 009/100 | Batch 350/600 | Cost: 0.0558\n",
      "Epoch: 009/100 | Batch 400/600 | Cost: 0.0566\n",
      "Epoch: 009/100 | Batch 450/600 | Cost: 0.0580\n",
      "Epoch: 009/100 | Batch 500/600 | Cost: 0.0507\n",
      "Epoch: 009/100 | Batch 550/600 | Cost: 0.0546\n",
      "Epoch: 009/100 Train Cost: 0.0532\n",
      "Time elapsed: 2.05 min\n",
      "Epoch: 010/100 | Batch 000/600 | Cost: 0.0503\n",
      "Epoch: 010/100 | Batch 050/600 | Cost: 0.0539\n",
      "Epoch: 010/100 | Batch 100/600 | Cost: 0.0519\n",
      "Epoch: 010/100 | Batch 150/600 | Cost: 0.0521\n",
      "Epoch: 010/100 | Batch 200/600 | Cost: 0.0555\n",
      "Epoch: 010/100 | Batch 250/600 | Cost: 0.0554\n",
      "Epoch: 010/100 | Batch 300/600 | Cost: 0.0472\n",
      "Epoch: 010/100 | Batch 350/600 | Cost: 0.0508\n",
      "Epoch: 010/100 | Batch 400/600 | Cost: 0.0482\n",
      "Epoch: 010/100 | Batch 450/600 | Cost: 0.0524\n",
      "Epoch: 010/100 | Batch 500/600 | Cost: 0.0519\n",
      "Epoch: 010/100 | Batch 550/600 | Cost: 0.0539\n",
      "Epoch: 010/100 Train Cost: 0.0502\n",
      "Time elapsed: 2.28 min\n",
      "Epoch: 011/100 | Batch 000/600 | Cost: 0.0488\n",
      "Epoch: 011/100 | Batch 050/600 | Cost: 0.0509\n",
      "Epoch: 011/100 | Batch 100/600 | Cost: 0.0457\n",
      "Epoch: 011/100 | Batch 150/600 | Cost: 0.0479\n",
      "Epoch: 011/100 | Batch 200/600 | Cost: 0.0493\n",
      "Epoch: 011/100 | Batch 250/600 | Cost: 0.0459\n",
      "Epoch: 011/100 | Batch 300/600 | Cost: 0.0457\n",
      "Epoch: 011/100 | Batch 350/600 | Cost: 0.0474\n",
      "Epoch: 011/100 | Batch 400/600 | Cost: 0.0487\n",
      "Epoch: 011/100 | Batch 450/600 | Cost: 0.0463\n",
      "Epoch: 011/100 | Batch 500/600 | Cost: 0.0504\n",
      "Epoch: 011/100 | Batch 550/600 | Cost: 0.0481\n",
      "Epoch: 011/100 Train Cost: 0.0475\n",
      "Time elapsed: 2.51 min\n",
      "Epoch: 012/100 | Batch 000/600 | Cost: 0.0429\n",
      "Epoch: 012/100 | Batch 050/600 | Cost: 0.0439\n",
      "Epoch: 012/100 | Batch 100/600 | Cost: 0.0438\n",
      "Epoch: 012/100 | Batch 150/600 | Cost: 0.0479\n",
      "Epoch: 012/100 | Batch 200/600 | Cost: 0.0442\n",
      "Epoch: 012/100 | Batch 250/600 | Cost: 0.0468\n",
      "Epoch: 012/100 | Batch 300/600 | Cost: 0.0424\n",
      "Epoch: 012/100 | Batch 350/600 | Cost: 0.0435\n",
      "Epoch: 012/100 | Batch 400/600 | Cost: 0.0474\n",
      "Epoch: 012/100 | Batch 450/600 | Cost: 0.0477\n",
      "Epoch: 012/100 | Batch 500/600 | Cost: 0.0466\n",
      "Epoch: 012/100 | Batch 550/600 | Cost: 0.0504\n",
      "Epoch: 012/100 Train Cost: 0.0452\n",
      "Time elapsed: 2.74 min\n",
      "Epoch: 013/100 | Batch 000/600 | Cost: 0.0427\n",
      "Epoch: 013/100 | Batch 050/600 | Cost: 0.0403\n",
      "Epoch: 013/100 | Batch 100/600 | Cost: 0.0445\n",
      "Epoch: 013/100 | Batch 150/600 | Cost: 0.0440\n",
      "Epoch: 013/100 | Batch 200/600 | Cost: 0.0462\n",
      "Epoch: 013/100 | Batch 250/600 | Cost: 0.0439\n",
      "Epoch: 013/100 | Batch 300/600 | Cost: 0.0445\n",
      "Epoch: 013/100 | Batch 350/600 | Cost: 0.0409\n",
      "Epoch: 013/100 | Batch 400/600 | Cost: 0.0439\n",
      "Epoch: 013/100 | Batch 450/600 | Cost: 0.0451\n",
      "Epoch: 013/100 | Batch 500/600 | Cost: 0.0421\n",
      "Epoch: 013/100 | Batch 550/600 | Cost: 0.0361\n",
      "Epoch: 013/100 Train Cost: 0.0432\n",
      "Time elapsed: 2.96 min\n",
      "Epoch: 014/100 | Batch 000/600 | Cost: 0.0457\n",
      "Epoch: 014/100 | Batch 050/600 | Cost: 0.0449\n",
      "Epoch: 014/100 | Batch 100/600 | Cost: 0.0468\n",
      "Epoch: 014/100 | Batch 150/600 | Cost: 0.0456\n",
      "Epoch: 014/100 | Batch 200/600 | Cost: 0.0387\n",
      "Epoch: 014/100 | Batch 250/600 | Cost: 0.0393\n",
      "Epoch: 014/100 | Batch 300/600 | Cost: 0.0420\n",
      "Epoch: 014/100 | Batch 350/600 | Cost: 0.0453\n",
      "Epoch: 014/100 | Batch 400/600 | Cost: 0.0415\n",
      "Epoch: 014/100 | Batch 450/600 | Cost: 0.0471\n",
      "Epoch: 014/100 | Batch 500/600 | Cost: 0.0477\n",
      "Epoch: 014/100 | Batch 550/600 | Cost: 0.0491\n",
      "Epoch: 014/100 Train Cost: 0.0414\n",
      "Time elapsed: 3.19 min\n",
      "Epoch: 015/100 | Batch 000/600 | Cost: 0.0430\n",
      "Epoch: 015/100 | Batch 050/600 | Cost: 0.0405\n",
      "Epoch: 015/100 | Batch 100/600 | Cost: 0.0437\n",
      "Epoch: 015/100 | Batch 150/600 | Cost: 0.0394\n",
      "Epoch: 015/100 | Batch 200/600 | Cost: 0.0398\n",
      "Epoch: 015/100 | Batch 250/600 | Cost: 0.0460\n",
      "Epoch: 015/100 | Batch 300/600 | Cost: 0.0376\n",
      "Epoch: 015/100 | Batch 350/600 | Cost: 0.0425\n",
      "Epoch: 015/100 | Batch 400/600 | Cost: 0.0405\n",
      "Epoch: 015/100 | Batch 450/600 | Cost: 0.0411\n",
      "Epoch: 015/100 | Batch 500/600 | Cost: 0.0417\n",
      "Epoch: 015/100 | Batch 550/600 | Cost: 0.0416\n",
      "Epoch: 015/100 Train Cost: 0.0398\n",
      "Time elapsed: 3.42 min\n",
      "Epoch: 016/100 | Batch 000/600 | Cost: 0.0425\n",
      "Epoch: 016/100 | Batch 050/600 | Cost: 0.0419\n",
      "Epoch: 016/100 | Batch 100/600 | Cost: 0.0402\n",
      "Epoch: 016/100 | Batch 150/600 | Cost: 0.0379\n",
      "Epoch: 016/100 | Batch 200/600 | Cost: 0.0347\n",
      "Epoch: 016/100 | Batch 250/600 | Cost: 0.0358\n",
      "Epoch: 016/100 | Batch 300/600 | Cost: 0.0384\n",
      "Epoch: 016/100 | Batch 350/600 | Cost: 0.0361\n",
      "Epoch: 016/100 | Batch 400/600 | Cost: 0.0411\n",
      "Epoch: 016/100 | Batch 450/600 | Cost: 0.0343\n",
      "Epoch: 016/100 | Batch 500/600 | Cost: 0.0387\n",
      "Epoch: 016/100 | Batch 550/600 | Cost: 0.0390\n",
      "Epoch: 016/100 Train Cost: 0.0384\n",
      "Time elapsed: 3.64 min\n",
      "Epoch: 017/100 | Batch 000/600 | Cost: 0.0381\n",
      "Epoch: 017/100 | Batch 050/600 | Cost: 0.0354\n",
      "Epoch: 017/100 | Batch 100/600 | Cost: 0.0386\n",
      "Epoch: 017/100 | Batch 150/600 | Cost: 0.0364\n",
      "Epoch: 017/100 | Batch 200/600 | Cost: 0.0320\n",
      "Epoch: 017/100 | Batch 250/600 | Cost: 0.0445\n",
      "Epoch: 017/100 | Batch 300/600 | Cost: 0.0422\n",
      "Epoch: 017/100 | Batch 350/600 | Cost: 0.0367\n",
      "Epoch: 017/100 | Batch 400/600 | Cost: 0.0389\n",
      "Epoch: 017/100 | Batch 450/600 | Cost: 0.0341\n",
      "Epoch: 017/100 | Batch 500/600 | Cost: 0.0383\n",
      "Epoch: 017/100 | Batch 550/600 | Cost: 0.0375\n",
      "Epoch: 017/100 Train Cost: 0.0372\n",
      "Time elapsed: 3.87 min\n",
      "Epoch: 018/100 | Batch 000/600 | Cost: 0.0395\n",
      "Epoch: 018/100 | Batch 050/600 | Cost: 0.0390\n",
      "Epoch: 018/100 | Batch 100/600 | Cost: 0.0417\n",
      "Epoch: 018/100 | Batch 150/600 | Cost: 0.0360\n",
      "Epoch: 018/100 | Batch 200/600 | Cost: 0.0378\n",
      "Epoch: 018/100 | Batch 250/600 | Cost: 0.0385\n",
      "Epoch: 018/100 | Batch 300/600 | Cost: 0.0374\n",
      "Epoch: 018/100 | Batch 350/600 | Cost: 0.0339\n",
      "Epoch: 018/100 | Batch 400/600 | Cost: 0.0387\n",
      "Epoch: 018/100 | Batch 450/600 | Cost: 0.0319\n",
      "Epoch: 018/100 | Batch 500/600 | Cost: 0.0358\n",
      "Epoch: 018/100 | Batch 550/600 | Cost: 0.0295\n",
      "Epoch: 018/100 Train Cost: 0.0360\n",
      "Time elapsed: 4.09 min\n",
      "Epoch: 019/100 | Batch 000/600 | Cost: 0.0429\n",
      "Epoch: 019/100 | Batch 050/600 | Cost: 0.0351\n",
      "Epoch: 019/100 | Batch 100/600 | Cost: 0.0386\n",
      "Epoch: 019/100 | Batch 150/600 | Cost: 0.0381\n",
      "Epoch: 019/100 | Batch 200/600 | Cost: 0.0305\n",
      "Epoch: 019/100 | Batch 250/600 | Cost: 0.0312\n",
      "Epoch: 019/100 | Batch 300/600 | Cost: 0.0301\n",
      "Epoch: 019/100 | Batch 350/600 | Cost: 0.0290\n",
      "Epoch: 019/100 | Batch 400/600 | Cost: 0.0361\n",
      "Epoch: 019/100 | Batch 450/600 | Cost: 0.0361\n",
      "Epoch: 019/100 | Batch 500/600 | Cost: 0.0350\n",
      "Epoch: 019/100 | Batch 550/600 | Cost: 0.0304\n",
      "Epoch: 019/100 Train Cost: 0.0350\n",
      "Time elapsed: 4.32 min\n",
      "Epoch: 020/100 | Batch 000/600 | Cost: 0.0338\n",
      "Epoch: 020/100 | Batch 050/600 | Cost: 0.0323\n",
      "Epoch: 020/100 | Batch 100/600 | Cost: 0.0352\n",
      "Epoch: 020/100 | Batch 150/600 | Cost: 0.0346\n",
      "Epoch: 020/100 | Batch 200/600 | Cost: 0.0348\n",
      "Epoch: 020/100 | Batch 250/600 | Cost: 0.0326\n",
      "Epoch: 020/100 | Batch 300/600 | Cost: 0.0352\n",
      "Epoch: 020/100 | Batch 350/600 | Cost: 0.0361\n",
      "Epoch: 020/100 | Batch 400/600 | Cost: 0.0386\n",
      "Epoch: 020/100 | Batch 450/600 | Cost: 0.0349\n",
      "Epoch: 020/100 | Batch 500/600 | Cost: 0.0378\n",
      "Epoch: 020/100 | Batch 550/600 | Cost: 0.0369\n",
      "Epoch: 020/100 Train Cost: 0.0340\n",
      "Time elapsed: 4.54 min\n",
      "Epoch: 021/100 | Batch 000/600 | Cost: 0.0300\n",
      "Epoch: 021/100 | Batch 050/600 | Cost: 0.0331\n",
      "Epoch: 021/100 | Batch 100/600 | Cost: 0.0289\n",
      "Epoch: 021/100 | Batch 150/600 | Cost: 0.0367\n",
      "Epoch: 021/100 | Batch 200/600 | Cost: 0.0351\n",
      "Epoch: 021/100 | Batch 250/600 | Cost: 0.0330\n",
      "Epoch: 021/100 | Batch 300/600 | Cost: 0.0359\n",
      "Epoch: 021/100 | Batch 350/600 | Cost: 0.0281\n",
      "Epoch: 021/100 | Batch 400/600 | Cost: 0.0336\n",
      "Epoch: 021/100 | Batch 450/600 | Cost: 0.0327\n",
      "Epoch: 021/100 | Batch 500/600 | Cost: 0.0304\n",
      "Epoch: 021/100 | Batch 550/600 | Cost: 0.0296\n",
      "Epoch: 021/100 Train Cost: 0.0331\n",
      "Time elapsed: 4.77 min\n",
      "Epoch: 022/100 | Batch 000/600 | Cost: 0.0322\n",
      "Epoch: 022/100 | Batch 050/600 | Cost: 0.0309\n",
      "Epoch: 022/100 | Batch 100/600 | Cost: 0.0326\n",
      "Epoch: 022/100 | Batch 150/600 | Cost: 0.0368\n",
      "Epoch: 022/100 | Batch 200/600 | Cost: 0.0317\n",
      "Epoch: 022/100 | Batch 250/600 | Cost: 0.0296\n",
      "Epoch: 022/100 | Batch 300/600 | Cost: 0.0327\n",
      "Epoch: 022/100 | Batch 350/600 | Cost: 0.0306\n",
      "Epoch: 022/100 | Batch 400/600 | Cost: 0.0260\n",
      "Epoch: 022/100 | Batch 450/600 | Cost: 0.0345\n",
      "Epoch: 022/100 | Batch 500/600 | Cost: 0.0327\n",
      "Epoch: 022/100 | Batch 550/600 | Cost: 0.0334\n",
      "Epoch: 022/100 Train Cost: 0.0323\n",
      "Time elapsed: 5.00 min\n",
      "Epoch: 023/100 | Batch 000/600 | Cost: 0.0286\n",
      "Epoch: 023/100 | Batch 050/600 | Cost: 0.0297\n",
      "Epoch: 023/100 | Batch 100/600 | Cost: 0.0314\n",
      "Epoch: 023/100 | Batch 150/600 | Cost: 0.0335\n",
      "Epoch: 023/100 | Batch 200/600 | Cost: 0.0326\n",
      "Epoch: 023/100 | Batch 250/600 | Cost: 0.0307\n",
      "Epoch: 023/100 | Batch 300/600 | Cost: 0.0339\n",
      "Epoch: 023/100 | Batch 350/600 | Cost: 0.0327\n",
      "Epoch: 023/100 | Batch 400/600 | Cost: 0.0386\n",
      "Epoch: 023/100 | Batch 450/600 | Cost: 0.0312\n",
      "Epoch: 023/100 | Batch 500/600 | Cost: 0.0282\n",
      "Epoch: 023/100 | Batch 550/600 | Cost: 0.0347\n",
      "Epoch: 023/100 Train Cost: 0.0316\n",
      "Time elapsed: 5.22 min\n",
      "Epoch: 024/100 | Batch 000/600 | Cost: 0.0364\n",
      "Epoch: 024/100 | Batch 050/600 | Cost: 0.0318\n",
      "Epoch: 024/100 | Batch 100/600 | Cost: 0.0309\n",
      "Epoch: 024/100 | Batch 150/600 | Cost: 0.0288\n",
      "Epoch: 024/100 | Batch 200/600 | Cost: 0.0296\n",
      "Epoch: 024/100 | Batch 250/600 | Cost: 0.0310\n",
      "Epoch: 024/100 | Batch 300/600 | Cost: 0.0354\n",
      "Epoch: 024/100 | Batch 350/600 | Cost: 0.0307\n",
      "Epoch: 024/100 | Batch 400/600 | Cost: 0.0344\n",
      "Epoch: 024/100 | Batch 450/600 | Cost: 0.0280\n",
      "Epoch: 024/100 | Batch 500/600 | Cost: 0.0261\n",
      "Epoch: 024/100 | Batch 550/600 | Cost: 0.0319\n",
      "Epoch: 024/100 Train Cost: 0.0309\n",
      "Time elapsed: 5.45 min\n",
      "Epoch: 025/100 | Batch 000/600 | Cost: 0.0260\n",
      "Epoch: 025/100 | Batch 050/600 | Cost: 0.0271\n",
      "Epoch: 025/100 | Batch 100/600 | Cost: 0.0302\n",
      "Epoch: 025/100 | Batch 150/600 | Cost: 0.0346\n",
      "Epoch: 025/100 | Batch 200/600 | Cost: 0.0318\n",
      "Epoch: 025/100 | Batch 250/600 | Cost: 0.0284\n",
      "Epoch: 025/100 | Batch 300/600 | Cost: 0.0290\n",
      "Epoch: 025/100 | Batch 350/600 | Cost: 0.0342\n",
      "Epoch: 025/100 | Batch 400/600 | Cost: 0.0293\n",
      "Epoch: 025/100 | Batch 450/600 | Cost: 0.0227\n",
      "Epoch: 025/100 | Batch 500/600 | Cost: 0.0268\n",
      "Epoch: 025/100 | Batch 550/600 | Cost: 0.0289\n",
      "Epoch: 025/100 Train Cost: 0.0303\n",
      "Time elapsed: 5.68 min\n",
      "Epoch: 026/100 | Batch 000/600 | Cost: 0.0328\n",
      "Epoch: 026/100 | Batch 050/600 | Cost: 0.0286\n",
      "Epoch: 026/100 | Batch 100/600 | Cost: 0.0279\n",
      "Epoch: 026/100 | Batch 150/600 | Cost: 0.0295\n",
      "Epoch: 026/100 | Batch 200/600 | Cost: 0.0256\n",
      "Epoch: 026/100 | Batch 250/600 | Cost: 0.0299\n",
      "Epoch: 026/100 | Batch 300/600 | Cost: 0.0325\n",
      "Epoch: 026/100 | Batch 350/600 | Cost: 0.0276\n",
      "Epoch: 026/100 | Batch 400/600 | Cost: 0.0340\n",
      "Epoch: 026/100 | Batch 450/600 | Cost: 0.0283\n",
      "Epoch: 026/100 | Batch 500/600 | Cost: 0.0345\n",
      "Epoch: 026/100 | Batch 550/600 | Cost: 0.0293\n",
      "Epoch: 026/100 Train Cost: 0.0297\n",
      "Time elapsed: 5.91 min\n",
      "Epoch: 027/100 | Batch 000/600 | Cost: 0.0271\n",
      "Epoch: 027/100 | Batch 050/600 | Cost: 0.0288\n",
      "Epoch: 027/100 | Batch 100/600 | Cost: 0.0342\n",
      "Epoch: 027/100 | Batch 150/600 | Cost: 0.0333\n",
      "Epoch: 027/100 | Batch 200/600 | Cost: 0.0319\n",
      "Epoch: 027/100 | Batch 250/600 | Cost: 0.0279\n",
      "Epoch: 027/100 | Batch 300/600 | Cost: 0.0281\n",
      "Epoch: 027/100 | Batch 350/600 | Cost: 0.0329\n",
      "Epoch: 027/100 | Batch 400/600 | Cost: 0.0343\n",
      "Epoch: 027/100 | Batch 450/600 | Cost: 0.0262\n",
      "Epoch: 027/100 | Batch 500/600 | Cost: 0.0271\n",
      "Epoch: 027/100 | Batch 550/600 | Cost: 0.0311\n",
      "Epoch: 027/100 Train Cost: 0.0291\n",
      "Time elapsed: 6.14 min\n",
      "Epoch: 028/100 | Batch 000/600 | Cost: 0.0311\n",
      "Epoch: 028/100 | Batch 050/600 | Cost: 0.0301\n",
      "Epoch: 028/100 | Batch 100/600 | Cost: 0.0268\n",
      "Epoch: 028/100 | Batch 150/600 | Cost: 0.0338\n",
      "Epoch: 028/100 | Batch 200/600 | Cost: 0.0315\n",
      "Epoch: 028/100 | Batch 250/600 | Cost: 0.0299\n",
      "Epoch: 028/100 | Batch 300/600 | Cost: 0.0280\n",
      "Epoch: 028/100 | Batch 350/600 | Cost: 0.0312\n",
      "Epoch: 028/100 | Batch 400/600 | Cost: 0.0308\n",
      "Epoch: 028/100 | Batch 450/600 | Cost: 0.0274\n",
      "Epoch: 028/100 | Batch 500/600 | Cost: 0.0285\n",
      "Epoch: 028/100 | Batch 550/600 | Cost: 0.0288\n",
      "Epoch: 028/100 Train Cost: 0.0286\n",
      "Time elapsed: 6.37 min\n",
      "Epoch: 029/100 | Batch 000/600 | Cost: 0.0241\n",
      "Epoch: 029/100 | Batch 050/600 | Cost: 0.0315\n",
      "Epoch: 029/100 | Batch 100/600 | Cost: 0.0292\n",
      "Epoch: 029/100 | Batch 150/600 | Cost: 0.0268\n",
      "Epoch: 029/100 | Batch 200/600 | Cost: 0.0287\n",
      "Epoch: 029/100 | Batch 250/600 | Cost: 0.0319\n",
      "Epoch: 029/100 | Batch 300/600 | Cost: 0.0276\n",
      "Epoch: 029/100 | Batch 350/600 | Cost: 0.0277\n",
      "Epoch: 029/100 | Batch 400/600 | Cost: 0.0282\n",
      "Epoch: 029/100 | Batch 450/600 | Cost: 0.0227\n",
      "Epoch: 029/100 | Batch 500/600 | Cost: 0.0345\n",
      "Epoch: 029/100 | Batch 550/600 | Cost: 0.0251\n",
      "Epoch: 029/100 Train Cost: 0.0281\n",
      "Time elapsed: 6.60 min\n",
      "Epoch: 030/100 | Batch 000/600 | Cost: 0.0308\n",
      "Epoch: 030/100 | Batch 050/600 | Cost: 0.0279\n",
      "Epoch: 030/100 | Batch 100/600 | Cost: 0.0249\n",
      "Epoch: 030/100 | Batch 150/600 | Cost: 0.0301\n",
      "Epoch: 030/100 | Batch 200/600 | Cost: 0.0322\n",
      "Epoch: 030/100 | Batch 250/600 | Cost: 0.0280\n",
      "Epoch: 030/100 | Batch 300/600 | Cost: 0.0299\n",
      "Epoch: 030/100 | Batch 350/600 | Cost: 0.0332\n",
      "Epoch: 030/100 | Batch 400/600 | Cost: 0.0246\n",
      "Epoch: 030/100 | Batch 450/600 | Cost: 0.0256\n",
      "Epoch: 030/100 | Batch 500/600 | Cost: 0.0316\n",
      "Epoch: 030/100 | Batch 550/600 | Cost: 0.0320\n",
      "Epoch: 030/100 Train Cost: 0.0276\n",
      "Time elapsed: 6.82 min\n",
      "Epoch: 031/100 | Batch 000/600 | Cost: 0.0269\n",
      "Epoch: 031/100 | Batch 050/600 | Cost: 0.0292\n",
      "Epoch: 031/100 | Batch 100/600 | Cost: 0.0343\n",
      "Epoch: 031/100 | Batch 150/600 | Cost: 0.0259\n",
      "Epoch: 031/100 | Batch 200/600 | Cost: 0.0257\n",
      "Epoch: 031/100 | Batch 250/600 | Cost: 0.0310\n",
      "Epoch: 031/100 | Batch 300/600 | Cost: 0.0261\n",
      "Epoch: 031/100 | Batch 350/600 | Cost: 0.0304\n",
      "Epoch: 031/100 | Batch 400/600 | Cost: 0.0316\n",
      "Epoch: 031/100 | Batch 450/600 | Cost: 0.0331\n",
      "Epoch: 031/100 | Batch 500/600 | Cost: 0.0302\n",
      "Epoch: 031/100 | Batch 550/600 | Cost: 0.0249\n",
      "Epoch: 031/100 Train Cost: 0.0272\n",
      "Time elapsed: 7.05 min\n",
      "Epoch: 032/100 | Batch 000/600 | Cost: 0.0304\n",
      "Epoch: 032/100 | Batch 050/600 | Cost: 0.0254\n",
      "Epoch: 032/100 | Batch 100/600 | Cost: 0.0270\n",
      "Epoch: 032/100 | Batch 150/600 | Cost: 0.0267\n",
      "Epoch: 032/100 | Batch 200/600 | Cost: 0.0312\n",
      "Epoch: 032/100 | Batch 250/600 | Cost: 0.0251\n",
      "Epoch: 032/100 | Batch 300/600 | Cost: 0.0251\n",
      "Epoch: 032/100 | Batch 350/600 | Cost: 0.0239\n",
      "Epoch: 032/100 | Batch 400/600 | Cost: 0.0219\n",
      "Epoch: 032/100 | Batch 450/600 | Cost: 0.0236\n",
      "Epoch: 032/100 | Batch 500/600 | Cost: 0.0258\n",
      "Epoch: 032/100 | Batch 550/600 | Cost: 0.0243\n",
      "Epoch: 032/100 Train Cost: 0.0268\n",
      "Time elapsed: 7.28 min\n",
      "Epoch: 033/100 | Batch 000/600 | Cost: 0.0296\n",
      "Epoch: 033/100 | Batch 050/600 | Cost: 0.0276\n",
      "Epoch: 033/100 | Batch 100/600 | Cost: 0.0243\n",
      "Epoch: 033/100 | Batch 150/600 | Cost: 0.0224\n",
      "Epoch: 033/100 | Batch 200/600 | Cost: 0.0304\n",
      "Epoch: 033/100 | Batch 250/600 | Cost: 0.0296\n",
      "Epoch: 033/100 | Batch 300/600 | Cost: 0.0333\n",
      "Epoch: 033/100 | Batch 350/600 | Cost: 0.0256\n",
      "Epoch: 033/100 | Batch 400/600 | Cost: 0.0192\n",
      "Epoch: 033/100 | Batch 450/600 | Cost: 0.0320\n",
      "Epoch: 033/100 | Batch 500/600 | Cost: 0.0273\n",
      "Epoch: 033/100 | Batch 550/600 | Cost: 0.0242\n",
      "Epoch: 033/100 Train Cost: 0.0264\n",
      "Time elapsed: 7.51 min\n",
      "Epoch: 034/100 | Batch 000/600 | Cost: 0.0227\n",
      "Epoch: 034/100 | Batch 050/600 | Cost: 0.0262\n",
      "Epoch: 034/100 | Batch 100/600 | Cost: 0.0274\n",
      "Epoch: 034/100 | Batch 150/600 | Cost: 0.0272\n",
      "Epoch: 034/100 | Batch 200/600 | Cost: 0.0242\n",
      "Epoch: 034/100 | Batch 250/600 | Cost: 0.0247\n",
      "Epoch: 034/100 | Batch 300/600 | Cost: 0.0247\n",
      "Epoch: 034/100 | Batch 350/600 | Cost: 0.0290\n",
      "Epoch: 034/100 | Batch 400/600 | Cost: 0.0281\n",
      "Epoch: 034/100 | Batch 450/600 | Cost: 0.0287\n",
      "Epoch: 034/100 | Batch 500/600 | Cost: 0.0283\n",
      "Epoch: 034/100 | Batch 550/600 | Cost: 0.0248\n",
      "Epoch: 034/100 Train Cost: 0.0260\n",
      "Time elapsed: 7.73 min\n",
      "Epoch: 035/100 | Batch 000/600 | Cost: 0.0218\n",
      "Epoch: 035/100 | Batch 050/600 | Cost: 0.0283\n",
      "Epoch: 035/100 | Batch 100/600 | Cost: 0.0245\n",
      "Epoch: 035/100 | Batch 150/600 | Cost: 0.0235\n",
      "Epoch: 035/100 | Batch 200/600 | Cost: 0.0251\n",
      "Epoch: 035/100 | Batch 250/600 | Cost: 0.0215\n",
      "Epoch: 035/100 | Batch 300/600 | Cost: 0.0322\n",
      "Epoch: 035/100 | Batch 350/600 | Cost: 0.0222\n",
      "Epoch: 035/100 | Batch 400/600 | Cost: 0.0265\n",
      "Epoch: 035/100 | Batch 450/600 | Cost: 0.0261\n",
      "Epoch: 035/100 | Batch 500/600 | Cost: 0.0261\n",
      "Epoch: 035/100 | Batch 550/600 | Cost: 0.0325\n",
      "Epoch: 035/100 Train Cost: 0.0257\n",
      "Time elapsed: 7.97 min\n",
      "Epoch: 036/100 | Batch 000/600 | Cost: 0.0255\n",
      "Epoch: 036/100 | Batch 050/600 | Cost: 0.0253\n",
      "Epoch: 036/100 | Batch 100/600 | Cost: 0.0191\n",
      "Epoch: 036/100 | Batch 150/600 | Cost: 0.0275\n",
      "Epoch: 036/100 | Batch 200/600 | Cost: 0.0224\n",
      "Epoch: 036/100 | Batch 250/600 | Cost: 0.0286\n",
      "Epoch: 036/100 | Batch 300/600 | Cost: 0.0327\n",
      "Epoch: 036/100 | Batch 350/600 | Cost: 0.0304\n",
      "Epoch: 036/100 | Batch 400/600 | Cost: 0.0299\n",
      "Epoch: 036/100 | Batch 450/600 | Cost: 0.0259\n",
      "Epoch: 036/100 | Batch 500/600 | Cost: 0.0242\n",
      "Epoch: 036/100 | Batch 550/600 | Cost: 0.0189\n",
      "Epoch: 036/100 Train Cost: 0.0253\n",
      "Time elapsed: 8.20 min\n",
      "Epoch: 037/100 | Batch 000/600 | Cost: 0.0276\n",
      "Epoch: 037/100 | Batch 050/600 | Cost: 0.0267\n",
      "Epoch: 037/100 | Batch 100/600 | Cost: 0.0234\n",
      "Epoch: 037/100 | Batch 150/600 | Cost: 0.0251\n",
      "Epoch: 037/100 | Batch 200/600 | Cost: 0.0271\n",
      "Epoch: 037/100 | Batch 250/600 | Cost: 0.0191\n",
      "Epoch: 037/100 | Batch 300/600 | Cost: 0.0232\n",
      "Epoch: 037/100 | Batch 350/600 | Cost: 0.0256\n",
      "Epoch: 037/100 | Batch 400/600 | Cost: 0.0239\n",
      "Epoch: 037/100 | Batch 450/600 | Cost: 0.0245\n",
      "Epoch: 037/100 | Batch 500/600 | Cost: 0.0202\n",
      "Epoch: 037/100 | Batch 550/600 | Cost: 0.0239\n",
      "Epoch: 037/100 Train Cost: 0.0250\n",
      "Time elapsed: 8.43 min\n",
      "Epoch: 038/100 | Batch 000/600 | Cost: 0.0289\n",
      "Epoch: 038/100 | Batch 050/600 | Cost: 0.0301\n",
      "Epoch: 038/100 | Batch 100/600 | Cost: 0.0231\n",
      "Epoch: 038/100 | Batch 150/600 | Cost: 0.0243\n",
      "Epoch: 038/100 | Batch 200/600 | Cost: 0.0198\n",
      "Epoch: 038/100 | Batch 250/600 | Cost: 0.0227\n",
      "Epoch: 038/100 | Batch 300/600 | Cost: 0.0224\n",
      "Epoch: 038/100 | Batch 350/600 | Cost: 0.0259\n",
      "Epoch: 038/100 | Batch 400/600 | Cost: 0.0270\n",
      "Epoch: 038/100 | Batch 450/600 | Cost: 0.0295\n",
      "Epoch: 038/100 | Batch 500/600 | Cost: 0.0262\n",
      "Epoch: 038/100 | Batch 550/600 | Cost: 0.0249\n",
      "Epoch: 038/100 Train Cost: 0.0247\n",
      "Time elapsed: 8.66 min\n",
      "Epoch: 039/100 | Batch 000/600 | Cost: 0.0289\n",
      "Epoch: 039/100 | Batch 050/600 | Cost: 0.0258\n",
      "Epoch: 039/100 | Batch 100/600 | Cost: 0.0230\n",
      "Epoch: 039/100 | Batch 150/600 | Cost: 0.0247\n",
      "Epoch: 039/100 | Batch 200/600 | Cost: 0.0253\n",
      "Epoch: 039/100 | Batch 250/600 | Cost: 0.0233\n",
      "Epoch: 039/100 | Batch 300/600 | Cost: 0.0228\n",
      "Epoch: 039/100 | Batch 350/600 | Cost: 0.0225\n",
      "Epoch: 039/100 | Batch 400/600 | Cost: 0.0254\n",
      "Epoch: 039/100 | Batch 450/600 | Cost: 0.0277\n",
      "Epoch: 039/100 | Batch 500/600 | Cost: 0.0283\n",
      "Epoch: 039/100 | Batch 550/600 | Cost: 0.0298\n",
      "Epoch: 039/100 Train Cost: 0.0244\n",
      "Time elapsed: 8.89 min\n",
      "Epoch: 040/100 | Batch 000/600 | Cost: 0.0205\n",
      "Epoch: 040/100 | Batch 050/600 | Cost: 0.0275\n",
      "Epoch: 040/100 | Batch 100/600 | Cost: 0.0251\n",
      "Epoch: 040/100 | Batch 150/600 | Cost: 0.0288\n",
      "Epoch: 040/100 | Batch 200/600 | Cost: 0.0238\n",
      "Epoch: 040/100 | Batch 250/600 | Cost: 0.0220\n",
      "Epoch: 040/100 | Batch 300/600 | Cost: 0.0233\n",
      "Epoch: 040/100 | Batch 350/600 | Cost: 0.0294\n",
      "Epoch: 040/100 | Batch 400/600 | Cost: 0.0269\n",
      "Epoch: 040/100 | Batch 450/600 | Cost: 0.0254\n",
      "Epoch: 040/100 | Batch 500/600 | Cost: 0.0240\n",
      "Epoch: 040/100 | Batch 550/600 | Cost: 0.0161\n",
      "Epoch: 040/100 Train Cost: 0.0242\n",
      "Time elapsed: 9.13 min\n",
      "Epoch: 041/100 | Batch 000/600 | Cost: 0.0254\n",
      "Epoch: 041/100 | Batch 050/600 | Cost: 0.0264\n",
      "Epoch: 041/100 | Batch 100/600 | Cost: 0.0229\n",
      "Epoch: 041/100 | Batch 150/600 | Cost: 0.0241\n",
      "Epoch: 041/100 | Batch 200/600 | Cost: 0.0234\n",
      "Epoch: 041/100 | Batch 250/600 | Cost: 0.0270\n",
      "Epoch: 041/100 | Batch 300/600 | Cost: 0.0266\n",
      "Epoch: 041/100 | Batch 350/600 | Cost: 0.0223\n",
      "Epoch: 041/100 | Batch 400/600 | Cost: 0.0236\n",
      "Epoch: 041/100 | Batch 450/600 | Cost: 0.0223\n",
      "Epoch: 041/100 | Batch 500/600 | Cost: 0.0254\n",
      "Epoch: 041/100 | Batch 550/600 | Cost: 0.0248\n",
      "Epoch: 041/100 Train Cost: 0.0239\n",
      "Time elapsed: 9.36 min\n",
      "Epoch: 042/100 | Batch 000/600 | Cost: 0.0243\n",
      "Epoch: 042/100 | Batch 050/600 | Cost: 0.0232\n",
      "Epoch: 042/100 | Batch 100/600 | Cost: 0.0272\n",
      "Epoch: 042/100 | Batch 150/600 | Cost: 0.0229\n",
      "Epoch: 042/100 | Batch 200/600 | Cost: 0.0224\n",
      "Epoch: 042/100 | Batch 250/600 | Cost: 0.0214\n",
      "Epoch: 042/100 | Batch 300/600 | Cost: 0.0226\n",
      "Epoch: 042/100 | Batch 350/600 | Cost: 0.0252\n",
      "Epoch: 042/100 | Batch 400/600 | Cost: 0.0145\n",
      "Epoch: 042/100 | Batch 450/600 | Cost: 0.0235\n",
      "Epoch: 042/100 | Batch 500/600 | Cost: 0.0245\n",
      "Epoch: 042/100 | Batch 550/600 | Cost: 0.0196\n",
      "Epoch: 042/100 Train Cost: 0.0237\n",
      "Time elapsed: 9.58 min\n",
      "Epoch: 043/100 | Batch 000/600 | Cost: 0.0260\n",
      "Epoch: 043/100 | Batch 050/600 | Cost: 0.0177\n",
      "Epoch: 043/100 | Batch 100/600 | Cost: 0.0239\n",
      "Epoch: 043/100 | Batch 150/600 | Cost: 0.0222\n",
      "Epoch: 043/100 | Batch 200/600 | Cost: 0.0213\n",
      "Epoch: 043/100 | Batch 250/600 | Cost: 0.0239\n",
      "Epoch: 043/100 | Batch 300/600 | Cost: 0.0191\n",
      "Epoch: 043/100 | Batch 350/600 | Cost: 0.0201\n",
      "Epoch: 043/100 | Batch 400/600 | Cost: 0.0251\n",
      "Epoch: 043/100 | Batch 450/600 | Cost: 0.0257\n",
      "Epoch: 043/100 | Batch 500/600 | Cost: 0.0218\n",
      "Epoch: 043/100 | Batch 550/600 | Cost: 0.0230\n",
      "Epoch: 043/100 Train Cost: 0.0234\n",
      "Time elapsed: 9.81 min\n",
      "Epoch: 044/100 | Batch 000/600 | Cost: 0.0288\n",
      "Epoch: 044/100 | Batch 050/600 | Cost: 0.0211\n",
      "Epoch: 044/100 | Batch 100/600 | Cost: 0.0277\n",
      "Epoch: 044/100 | Batch 150/600 | Cost: 0.0198\n",
      "Epoch: 044/100 | Batch 200/600 | Cost: 0.0239\n",
      "Epoch: 044/100 | Batch 250/600 | Cost: 0.0190\n",
      "Epoch: 044/100 | Batch 300/600 | Cost: 0.0249\n",
      "Epoch: 044/100 | Batch 350/600 | Cost: 0.0218\n",
      "Epoch: 044/100 | Batch 400/600 | Cost: 0.0264\n",
      "Epoch: 044/100 | Batch 450/600 | Cost: 0.0265\n",
      "Epoch: 044/100 | Batch 500/600 | Cost: 0.0212\n",
      "Epoch: 044/100 | Batch 550/600 | Cost: 0.0218\n",
      "Epoch: 044/100 Train Cost: 0.0232\n",
      "Time elapsed: 10.03 min\n",
      "Epoch: 045/100 | Batch 000/600 | Cost: 0.0228\n",
      "Epoch: 045/100 | Batch 050/600 | Cost: 0.0236\n",
      "Epoch: 045/100 | Batch 100/600 | Cost: 0.0235\n",
      "Epoch: 045/100 | Batch 150/600 | Cost: 0.0198\n",
      "Epoch: 045/100 | Batch 200/600 | Cost: 0.0217\n",
      "Epoch: 045/100 | Batch 250/600 | Cost: 0.0242\n",
      "Epoch: 045/100 | Batch 300/600 | Cost: 0.0246\n",
      "Epoch: 045/100 | Batch 350/600 | Cost: 0.0232\n",
      "Epoch: 045/100 | Batch 400/600 | Cost: 0.0279\n",
      "Epoch: 045/100 | Batch 450/600 | Cost: 0.0261\n",
      "Epoch: 045/100 | Batch 500/600 | Cost: 0.0332\n",
      "Epoch: 045/100 | Batch 550/600 | Cost: 0.0224\n",
      "Epoch: 045/100 Train Cost: 0.0230\n",
      "Time elapsed: 10.26 min\n",
      "Epoch: 046/100 | Batch 000/600 | Cost: 0.0255\n",
      "Epoch: 046/100 | Batch 050/600 | Cost: 0.0232\n",
      "Epoch: 046/100 | Batch 100/600 | Cost: 0.0172\n",
      "Epoch: 046/100 | Batch 150/600 | Cost: 0.0188\n",
      "Epoch: 046/100 | Batch 200/600 | Cost: 0.0215\n",
      "Epoch: 046/100 | Batch 250/600 | Cost: 0.0229\n",
      "Epoch: 046/100 | Batch 300/600 | Cost: 0.0182\n",
      "Epoch: 046/100 | Batch 350/600 | Cost: 0.0232\n",
      "Epoch: 046/100 | Batch 400/600 | Cost: 0.0228\n",
      "Epoch: 046/100 | Batch 450/600 | Cost: 0.0211\n",
      "Epoch: 046/100 | Batch 500/600 | Cost: 0.0220\n",
      "Epoch: 046/100 | Batch 550/600 | Cost: 0.0217\n",
      "Epoch: 046/100 Train Cost: 0.0228\n",
      "Time elapsed: 10.49 min\n",
      "Epoch: 047/100 | Batch 000/600 | Cost: 0.0223\n",
      "Epoch: 047/100 | Batch 050/600 | Cost: 0.0238\n",
      "Epoch: 047/100 | Batch 100/600 | Cost: 0.0218\n",
      "Epoch: 047/100 | Batch 150/600 | Cost: 0.0250\n",
      "Epoch: 047/100 | Batch 200/600 | Cost: 0.0213\n",
      "Epoch: 047/100 | Batch 250/600 | Cost: 0.0191\n",
      "Epoch: 047/100 | Batch 300/600 | Cost: 0.0265\n",
      "Epoch: 047/100 | Batch 350/600 | Cost: 0.0192\n",
      "Epoch: 047/100 | Batch 400/600 | Cost: 0.0212\n",
      "Epoch: 047/100 | Batch 450/600 | Cost: 0.0220\n",
      "Epoch: 047/100 | Batch 500/600 | Cost: 0.0231\n",
      "Epoch: 047/100 | Batch 550/600 | Cost: 0.0250\n",
      "Epoch: 047/100 Train Cost: 0.0226\n",
      "Time elapsed: 10.72 min\n",
      "Epoch: 048/100 | Batch 000/600 | Cost: 0.0219\n",
      "Epoch: 048/100 | Batch 050/600 | Cost: 0.0217\n",
      "Epoch: 048/100 | Batch 100/600 | Cost: 0.0258\n",
      "Epoch: 048/100 | Batch 150/600 | Cost: 0.0236\n",
      "Epoch: 048/100 | Batch 200/600 | Cost: 0.0244\n",
      "Epoch: 048/100 | Batch 250/600 | Cost: 0.0151\n",
      "Epoch: 048/100 | Batch 300/600 | Cost: 0.0155\n",
      "Epoch: 048/100 | Batch 350/600 | Cost: 0.0305\n",
      "Epoch: 048/100 | Batch 400/600 | Cost: 0.0225\n",
      "Epoch: 048/100 | Batch 450/600 | Cost: 0.0237\n",
      "Epoch: 048/100 | Batch 500/600 | Cost: 0.0244\n",
      "Epoch: 048/100 | Batch 550/600 | Cost: 0.0234\n",
      "Epoch: 048/100 Train Cost: 0.0224\n",
      "Time elapsed: 10.95 min\n",
      "Epoch: 049/100 | Batch 000/600 | Cost: 0.0245\n",
      "Epoch: 049/100 | Batch 050/600 | Cost: 0.0233\n",
      "Epoch: 049/100 | Batch 100/600 | Cost: 0.0239\n",
      "Epoch: 049/100 | Batch 150/600 | Cost: 0.0225\n",
      "Epoch: 049/100 | Batch 200/600 | Cost: 0.0157\n",
      "Epoch: 049/100 | Batch 250/600 | Cost: 0.0249\n",
      "Epoch: 049/100 | Batch 300/600 | Cost: 0.0227\n",
      "Epoch: 049/100 | Batch 350/600 | Cost: 0.0237\n",
      "Epoch: 049/100 | Batch 400/600 | Cost: 0.0278\n",
      "Epoch: 049/100 | Batch 450/600 | Cost: 0.0247\n",
      "Epoch: 049/100 | Batch 500/600 | Cost: 0.0173\n",
      "Epoch: 049/100 | Batch 550/600 | Cost: 0.0181\n",
      "Epoch: 049/100 Train Cost: 0.0222\n",
      "Time elapsed: 11.17 min\n",
      "Epoch: 050/100 | Batch 000/600 | Cost: 0.0218\n",
      "Epoch: 050/100 | Batch 050/600 | Cost: 0.0188\n",
      "Epoch: 050/100 | Batch 100/600 | Cost: 0.0209\n",
      "Epoch: 050/100 | Batch 150/600 | Cost: 0.0181\n",
      "Epoch: 050/100 | Batch 200/600 | Cost: 0.0270\n",
      "Epoch: 050/100 | Batch 250/600 | Cost: 0.0224\n",
      "Epoch: 050/100 | Batch 300/600 | Cost: 0.0218\n",
      "Epoch: 050/100 | Batch 350/600 | Cost: 0.0290\n",
      "Epoch: 050/100 | Batch 400/600 | Cost: 0.0263\n",
      "Epoch: 050/100 | Batch 450/600 | Cost: 0.0194\n",
      "Epoch: 050/100 | Batch 500/600 | Cost: 0.0152\n",
      "Epoch: 050/100 | Batch 550/600 | Cost: 0.0182\n",
      "Epoch: 050/100 Train Cost: 0.0220\n",
      "Time elapsed: 11.41 min\n",
      "Epoch: 051/100 | Batch 000/600 | Cost: 0.0230\n",
      "Epoch: 051/100 | Batch 050/600 | Cost: 0.0244\n",
      "Epoch: 051/100 | Batch 100/600 | Cost: 0.0171\n",
      "Epoch: 051/100 | Batch 150/600 | Cost: 0.0232\n",
      "Epoch: 051/100 | Batch 200/600 | Cost: 0.0197\n",
      "Epoch: 051/100 | Batch 250/600 | Cost: 0.0206\n",
      "Epoch: 051/100 | Batch 300/600 | Cost: 0.0174\n",
      "Epoch: 051/100 | Batch 350/600 | Cost: 0.0226\n",
      "Epoch: 051/100 | Batch 400/600 | Cost: 0.0202\n",
      "Epoch: 051/100 | Batch 450/600 | Cost: 0.0220\n",
      "Epoch: 051/100 | Batch 500/600 | Cost: 0.0252\n",
      "Epoch: 051/100 | Batch 550/600 | Cost: 0.0174\n",
      "Epoch: 051/100 Train Cost: 0.0218\n",
      "Time elapsed: 11.64 min\n",
      "Epoch: 052/100 | Batch 000/600 | Cost: 0.0177\n",
      "Epoch: 052/100 | Batch 050/600 | Cost: 0.0222\n",
      "Epoch: 052/100 | Batch 100/600 | Cost: 0.0197\n",
      "Epoch: 052/100 | Batch 150/600 | Cost: 0.0140\n",
      "Epoch: 052/100 | Batch 200/600 | Cost: 0.0245\n",
      "Epoch: 052/100 | Batch 250/600 | Cost: 0.0232\n",
      "Epoch: 052/100 | Batch 300/600 | Cost: 0.0227\n",
      "Epoch: 052/100 | Batch 350/600 | Cost: 0.0229\n",
      "Epoch: 052/100 | Batch 400/600 | Cost: 0.0244\n",
      "Epoch: 052/100 | Batch 450/600 | Cost: 0.0268\n",
      "Epoch: 052/100 | Batch 500/600 | Cost: 0.0208\n",
      "Epoch: 052/100 | Batch 550/600 | Cost: 0.0196\n",
      "Epoch: 052/100 Train Cost: 0.0217\n",
      "Time elapsed: 11.86 min\n",
      "Epoch: 053/100 | Batch 000/600 | Cost: 0.0272\n",
      "Epoch: 053/100 | Batch 050/600 | Cost: 0.0201\n",
      "Epoch: 053/100 | Batch 100/600 | Cost: 0.0214\n",
      "Epoch: 053/100 | Batch 150/600 | Cost: 0.0234\n",
      "Epoch: 053/100 | Batch 200/600 | Cost: 0.0216\n",
      "Epoch: 053/100 | Batch 250/600 | Cost: 0.0224\n",
      "Epoch: 053/100 | Batch 300/600 | Cost: 0.0211\n",
      "Epoch: 053/100 | Batch 350/600 | Cost: 0.0284\n",
      "Epoch: 053/100 | Batch 400/600 | Cost: 0.0197\n",
      "Epoch: 053/100 | Batch 450/600 | Cost: 0.0189\n",
      "Epoch: 053/100 | Batch 500/600 | Cost: 0.0262\n",
      "Epoch: 053/100 | Batch 550/600 | Cost: 0.0242\n",
      "Epoch: 053/100 Train Cost: 0.0215\n",
      "Time elapsed: 12.09 min\n",
      "Epoch: 054/100 | Batch 000/600 | Cost: 0.0168\n",
      "Epoch: 054/100 | Batch 050/600 | Cost: 0.0208\n",
      "Epoch: 054/100 | Batch 100/600 | Cost: 0.0261\n",
      "Epoch: 054/100 | Batch 150/600 | Cost: 0.0205\n",
      "Epoch: 054/100 | Batch 200/600 | Cost: 0.0212\n",
      "Epoch: 054/100 | Batch 250/600 | Cost: 0.0228\n",
      "Epoch: 054/100 | Batch 300/600 | Cost: 0.0202\n",
      "Epoch: 054/100 | Batch 350/600 | Cost: 0.0194\n",
      "Epoch: 054/100 | Batch 400/600 | Cost: 0.0210\n",
      "Epoch: 054/100 | Batch 450/600 | Cost: 0.0240\n",
      "Epoch: 054/100 | Batch 500/600 | Cost: 0.0193\n",
      "Epoch: 054/100 | Batch 550/600 | Cost: 0.0220\n",
      "Epoch: 054/100 Train Cost: 0.0214\n",
      "Time elapsed: 12.32 min\n",
      "Epoch: 055/100 | Batch 000/600 | Cost: 0.0264\n",
      "Epoch: 055/100 | Batch 050/600 | Cost: 0.0203\n",
      "Epoch: 055/100 | Batch 100/600 | Cost: 0.0192\n",
      "Epoch: 055/100 | Batch 150/600 | Cost: 0.0208\n",
      "Epoch: 055/100 | Batch 200/600 | Cost: 0.0251\n",
      "Epoch: 055/100 | Batch 250/600 | Cost: 0.0254\n",
      "Epoch: 055/100 | Batch 300/600 | Cost: 0.0178\n",
      "Epoch: 055/100 | Batch 350/600 | Cost: 0.0207\n",
      "Epoch: 055/100 | Batch 400/600 | Cost: 0.0243\n",
      "Epoch: 055/100 | Batch 450/600 | Cost: 0.0191\n",
      "Epoch: 055/100 | Batch 500/600 | Cost: 0.0174\n",
      "Epoch: 055/100 | Batch 550/600 | Cost: 0.0221\n",
      "Epoch: 055/100 Train Cost: 0.0212\n",
      "Time elapsed: 12.55 min\n",
      "Epoch: 056/100 | Batch 000/600 | Cost: 0.0177\n",
      "Epoch: 056/100 | Batch 050/600 | Cost: 0.0188\n",
      "Epoch: 056/100 | Batch 100/600 | Cost: 0.0205\n",
      "Epoch: 056/100 | Batch 150/600 | Cost: 0.0215\n",
      "Epoch: 056/100 | Batch 200/600 | Cost: 0.0212\n",
      "Epoch: 056/100 | Batch 250/600 | Cost: 0.0220\n",
      "Epoch: 056/100 | Batch 300/600 | Cost: 0.0238\n",
      "Epoch: 056/100 | Batch 350/600 | Cost: 0.0196\n",
      "Epoch: 056/100 | Batch 400/600 | Cost: 0.0205\n",
      "Epoch: 056/100 | Batch 450/600 | Cost: 0.0249\n",
      "Epoch: 056/100 | Batch 500/600 | Cost: 0.0234\n",
      "Epoch: 056/100 | Batch 550/600 | Cost: 0.0196\n",
      "Epoch: 056/100 Train Cost: 0.0211\n",
      "Time elapsed: 12.78 min\n",
      "Epoch: 057/100 | Batch 000/600 | Cost: 0.0184\n",
      "Epoch: 057/100 | Batch 050/600 | Cost: 0.0202\n",
      "Epoch: 057/100 | Batch 100/600 | Cost: 0.0236\n",
      "Epoch: 057/100 | Batch 150/600 | Cost: 0.0230\n",
      "Epoch: 057/100 | Batch 200/600 | Cost: 0.0183\n",
      "Epoch: 057/100 | Batch 250/600 | Cost: 0.0175\n",
      "Epoch: 057/100 | Batch 300/600 | Cost: 0.0203\n",
      "Epoch: 057/100 | Batch 350/600 | Cost: 0.0162\n",
      "Epoch: 057/100 | Batch 400/600 | Cost: 0.0258\n",
      "Epoch: 057/100 | Batch 450/600 | Cost: 0.0186\n",
      "Epoch: 057/100 | Batch 500/600 | Cost: 0.0214\n",
      "Epoch: 057/100 | Batch 550/600 | Cost: 0.0199\n",
      "Epoch: 057/100 Train Cost: 0.0209\n",
      "Time elapsed: 13.01 min\n",
      "Epoch: 058/100 | Batch 000/600 | Cost: 0.0206\n",
      "Epoch: 058/100 | Batch 050/600 | Cost: 0.0164\n",
      "Epoch: 058/100 | Batch 100/600 | Cost: 0.0215\n",
      "Epoch: 058/100 | Batch 150/600 | Cost: 0.0232\n",
      "Epoch: 058/100 | Batch 200/600 | Cost: 0.0192\n",
      "Epoch: 058/100 | Batch 250/600 | Cost: 0.0217\n",
      "Epoch: 058/100 | Batch 300/600 | Cost: 0.0282\n",
      "Epoch: 058/100 | Batch 350/600 | Cost: 0.0188\n",
      "Epoch: 058/100 | Batch 400/600 | Cost: 0.0218\n",
      "Epoch: 058/100 | Batch 450/600 | Cost: 0.0235\n",
      "Epoch: 058/100 | Batch 500/600 | Cost: 0.0245\n",
      "Epoch: 058/100 | Batch 550/600 | Cost: 0.0202\n",
      "Epoch: 058/100 Train Cost: 0.0208\n",
      "Time elapsed: 13.24 min\n",
      "Epoch: 059/100 | Batch 000/600 | Cost: 0.0176\n",
      "Epoch: 059/100 | Batch 050/600 | Cost: 0.0185\n",
      "Epoch: 059/100 | Batch 100/600 | Cost: 0.0184\n",
      "Epoch: 059/100 | Batch 150/600 | Cost: 0.0216\n",
      "Epoch: 059/100 | Batch 200/600 | Cost: 0.0208\n",
      "Epoch: 059/100 | Batch 250/600 | Cost: 0.0178\n",
      "Epoch: 059/100 | Batch 300/600 | Cost: 0.0138\n",
      "Epoch: 059/100 | Batch 350/600 | Cost: 0.0212\n",
      "Epoch: 059/100 | Batch 400/600 | Cost: 0.0296\n",
      "Epoch: 059/100 | Batch 450/600 | Cost: 0.0158\n",
      "Epoch: 059/100 | Batch 500/600 | Cost: 0.0191\n",
      "Epoch: 059/100 | Batch 550/600 | Cost: 0.0245\n",
      "Epoch: 059/100 Train Cost: 0.0207\n",
      "Time elapsed: 13.46 min\n",
      "Epoch: 060/100 | Batch 000/600 | Cost: 0.0215\n",
      "Epoch: 060/100 | Batch 050/600 | Cost: 0.0229\n",
      "Epoch: 060/100 | Batch 100/600 | Cost: 0.0229\n",
      "Epoch: 060/100 | Batch 150/600 | Cost: 0.0153\n",
      "Epoch: 060/100 | Batch 200/600 | Cost: 0.0239\n",
      "Epoch: 060/100 | Batch 250/600 | Cost: 0.0237\n",
      "Epoch: 060/100 | Batch 300/600 | Cost: 0.0204\n",
      "Epoch: 060/100 | Batch 350/600 | Cost: 0.0235\n",
      "Epoch: 060/100 | Batch 400/600 | Cost: 0.0245\n",
      "Epoch: 060/100 | Batch 450/600 | Cost: 0.0222\n",
      "Epoch: 060/100 | Batch 500/600 | Cost: 0.0230\n",
      "Epoch: 060/100 | Batch 550/600 | Cost: 0.0194\n",
      "Epoch: 060/100 Train Cost: 0.0205\n",
      "Time elapsed: 13.69 min\n",
      "Epoch: 061/100 | Batch 000/600 | Cost: 0.0235\n",
      "Epoch: 061/100 | Batch 050/600 | Cost: 0.0252\n",
      "Epoch: 061/100 | Batch 100/600 | Cost: 0.0247\n",
      "Epoch: 061/100 | Batch 150/600 | Cost: 0.0199\n",
      "Epoch: 061/100 | Batch 200/600 | Cost: 0.0207\n",
      "Epoch: 061/100 | Batch 250/600 | Cost: 0.0261\n",
      "Epoch: 061/100 | Batch 300/600 | Cost: 0.0255\n",
      "Epoch: 061/100 | Batch 350/600 | Cost: 0.0145\n",
      "Epoch: 061/100 | Batch 400/600 | Cost: 0.0200\n",
      "Epoch: 061/100 | Batch 450/600 | Cost: 0.0194\n",
      "Epoch: 061/100 | Batch 500/600 | Cost: 0.0220\n",
      "Epoch: 061/100 | Batch 550/600 | Cost: 0.0203\n",
      "Epoch: 061/100 Train Cost: 0.0204\n",
      "Time elapsed: 13.92 min\n",
      "Epoch: 062/100 | Batch 000/600 | Cost: 0.0179\n",
      "Epoch: 062/100 | Batch 050/600 | Cost: 0.0196\n",
      "Epoch: 062/100 | Batch 100/600 | Cost: 0.0186\n",
      "Epoch: 062/100 | Batch 150/600 | Cost: 0.0136\n",
      "Epoch: 062/100 | Batch 200/600 | Cost: 0.0168\n",
      "Epoch: 062/100 | Batch 250/600 | Cost: 0.0224\n",
      "Epoch: 062/100 | Batch 300/600 | Cost: 0.0240\n",
      "Epoch: 062/100 | Batch 350/600 | Cost: 0.0231\n",
      "Epoch: 062/100 | Batch 400/600 | Cost: 0.0202\n",
      "Epoch: 062/100 | Batch 450/600 | Cost: 0.0187\n",
      "Epoch: 062/100 | Batch 500/600 | Cost: 0.0162\n",
      "Epoch: 062/100 | Batch 550/600 | Cost: 0.0281\n",
      "Epoch: 062/100 Train Cost: 0.0203\n",
      "Time elapsed: 14.15 min\n",
      "Epoch: 063/100 | Batch 000/600 | Cost: 0.0244\n",
      "Epoch: 063/100 | Batch 050/600 | Cost: 0.0210\n",
      "Epoch: 063/100 | Batch 100/600 | Cost: 0.0190\n",
      "Epoch: 063/100 | Batch 150/600 | Cost: 0.0227\n",
      "Epoch: 063/100 | Batch 200/600 | Cost: 0.0205\n",
      "Epoch: 063/100 | Batch 250/600 | Cost: 0.0252\n",
      "Epoch: 063/100 | Batch 300/600 | Cost: 0.0227\n",
      "Epoch: 063/100 | Batch 350/600 | Cost: 0.0198\n",
      "Epoch: 063/100 | Batch 400/600 | Cost: 0.0228\n",
      "Epoch: 063/100 | Batch 450/600 | Cost: 0.0153\n",
      "Epoch: 063/100 | Batch 500/600 | Cost: 0.0172\n",
      "Epoch: 063/100 | Batch 550/600 | Cost: 0.0187\n",
      "Epoch: 063/100 Train Cost: 0.0202\n",
      "Time elapsed: 14.38 min\n",
      "Epoch: 064/100 | Batch 000/600 | Cost: 0.0163\n",
      "Epoch: 064/100 | Batch 050/600 | Cost: 0.0231\n",
      "Epoch: 064/100 | Batch 100/600 | Cost: 0.0194\n",
      "Epoch: 064/100 | Batch 150/600 | Cost: 0.0211\n",
      "Epoch: 064/100 | Batch 200/600 | Cost: 0.0230\n",
      "Epoch: 064/100 | Batch 250/600 | Cost: 0.0256\n",
      "Epoch: 064/100 | Batch 300/600 | Cost: 0.0179\n",
      "Epoch: 064/100 | Batch 350/600 | Cost: 0.0168\n",
      "Epoch: 064/100 | Batch 400/600 | Cost: 0.0260\n",
      "Epoch: 064/100 | Batch 450/600 | Cost: 0.0224\n",
      "Epoch: 064/100 | Batch 500/600 | Cost: 0.0160\n",
      "Epoch: 064/100 | Batch 550/600 | Cost: 0.0211\n",
      "Epoch: 064/100 Train Cost: 0.0200\n",
      "Time elapsed: 14.61 min\n",
      "Epoch: 065/100 | Batch 000/600 | Cost: 0.0224\n",
      "Epoch: 065/100 | Batch 050/600 | Cost: 0.0182\n",
      "Epoch: 065/100 | Batch 100/600 | Cost: 0.0220\n",
      "Epoch: 065/100 | Batch 150/600 | Cost: 0.0212\n",
      "Epoch: 065/100 | Batch 200/600 | Cost: 0.0201\n",
      "Epoch: 065/100 | Batch 250/600 | Cost: 0.0185\n",
      "Epoch: 065/100 | Batch 300/600 | Cost: 0.0243\n",
      "Epoch: 065/100 | Batch 350/600 | Cost: 0.0230\n",
      "Epoch: 065/100 | Batch 400/600 | Cost: 0.0224\n",
      "Epoch: 065/100 | Batch 450/600 | Cost: 0.0155\n",
      "Epoch: 065/100 | Batch 500/600 | Cost: 0.0180\n",
      "Epoch: 065/100 | Batch 550/600 | Cost: 0.0231\n",
      "Epoch: 065/100 Train Cost: 0.0199\n",
      "Time elapsed: 14.84 min\n",
      "Epoch: 066/100 | Batch 000/600 | Cost: 0.0206\n",
      "Epoch: 066/100 | Batch 050/600 | Cost: 0.0214\n",
      "Epoch: 066/100 | Batch 100/600 | Cost: 0.0186\n",
      "Epoch: 066/100 | Batch 150/600 | Cost: 0.0215\n",
      "Epoch: 066/100 | Batch 200/600 | Cost: 0.0234\n",
      "Epoch: 066/100 | Batch 250/600 | Cost: 0.0221\n",
      "Epoch: 066/100 | Batch 300/600 | Cost: 0.0211\n",
      "Epoch: 066/100 | Batch 350/600 | Cost: 0.0174\n",
      "Epoch: 066/100 | Batch 400/600 | Cost: 0.0169\n",
      "Epoch: 066/100 | Batch 450/600 | Cost: 0.0169\n",
      "Epoch: 066/100 | Batch 500/600 | Cost: 0.0203\n",
      "Epoch: 066/100 | Batch 550/600 | Cost: 0.0192\n",
      "Epoch: 066/100 Train Cost: 0.0198\n",
      "Time elapsed: 15.07 min\n",
      "Epoch: 067/100 | Batch 000/600 | Cost: 0.0189\n",
      "Epoch: 067/100 | Batch 050/600 | Cost: 0.0201\n",
      "Epoch: 067/100 | Batch 100/600 | Cost: 0.0184\n",
      "Epoch: 067/100 | Batch 150/600 | Cost: 0.0181\n",
      "Epoch: 067/100 | Batch 200/600 | Cost: 0.0198\n",
      "Epoch: 067/100 | Batch 250/600 | Cost: 0.0197\n",
      "Epoch: 067/100 | Batch 300/600 | Cost: 0.0244\n",
      "Epoch: 067/100 | Batch 350/600 | Cost: 0.0162\n",
      "Epoch: 067/100 | Batch 400/600 | Cost: 0.0222\n",
      "Epoch: 067/100 | Batch 450/600 | Cost: 0.0168\n",
      "Epoch: 067/100 | Batch 500/600 | Cost: 0.0154\n",
      "Epoch: 067/100 | Batch 550/600 | Cost: 0.0269\n",
      "Epoch: 067/100 Train Cost: 0.0197\n",
      "Time elapsed: 15.30 min\n",
      "Epoch: 068/100 | Batch 000/600 | Cost: 0.0223\n",
      "Epoch: 068/100 | Batch 050/600 | Cost: 0.0200\n",
      "Epoch: 068/100 | Batch 100/600 | Cost: 0.0230\n",
      "Epoch: 068/100 | Batch 150/600 | Cost: 0.0214\n",
      "Epoch: 068/100 | Batch 200/600 | Cost: 0.0200\n",
      "Epoch: 068/100 | Batch 250/600 | Cost: 0.0205\n",
      "Epoch: 068/100 | Batch 300/600 | Cost: 0.0214\n",
      "Epoch: 068/100 | Batch 350/600 | Cost: 0.0131\n",
      "Epoch: 068/100 | Batch 400/600 | Cost: 0.0249\n",
      "Epoch: 068/100 | Batch 450/600 | Cost: 0.0198\n",
      "Epoch: 068/100 | Batch 500/600 | Cost: 0.0172\n",
      "Epoch: 068/100 | Batch 550/600 | Cost: 0.0186\n",
      "Epoch: 068/100 Train Cost: 0.0196\n",
      "Time elapsed: 15.52 min\n",
      "Epoch: 069/100 | Batch 000/600 | Cost: 0.0226\n",
      "Epoch: 069/100 | Batch 050/600 | Cost: 0.0171\n",
      "Epoch: 069/100 | Batch 100/600 | Cost: 0.0223\n",
      "Epoch: 069/100 | Batch 150/600 | Cost: 0.0170\n",
      "Epoch: 069/100 | Batch 200/600 | Cost: 0.0210\n",
      "Epoch: 069/100 | Batch 250/600 | Cost: 0.0214\n",
      "Epoch: 069/100 | Batch 300/600 | Cost: 0.0239\n",
      "Epoch: 069/100 | Batch 350/600 | Cost: 0.0230\n",
      "Epoch: 069/100 | Batch 400/600 | Cost: 0.0200\n",
      "Epoch: 069/100 | Batch 450/600 | Cost: 0.0226\n",
      "Epoch: 069/100 | Batch 500/600 | Cost: 0.0174\n",
      "Epoch: 069/100 | Batch 550/600 | Cost: 0.0205\n",
      "Epoch: 069/100 Train Cost: 0.0195\n",
      "Time elapsed: 15.75 min\n",
      "Epoch: 070/100 | Batch 000/600 | Cost: 0.0185\n",
      "Epoch: 070/100 | Batch 050/600 | Cost: 0.0152\n",
      "Epoch: 070/100 | Batch 100/600 | Cost: 0.0260\n",
      "Epoch: 070/100 | Batch 150/600 | Cost: 0.0192\n",
      "Epoch: 070/100 | Batch 200/600 | Cost: 0.0192\n",
      "Epoch: 070/100 | Batch 250/600 | Cost: 0.0143\n",
      "Epoch: 070/100 | Batch 300/600 | Cost: 0.0207\n",
      "Epoch: 070/100 | Batch 350/600 | Cost: 0.0233\n",
      "Epoch: 070/100 | Batch 400/600 | Cost: 0.0190\n",
      "Epoch: 070/100 | Batch 450/600 | Cost: 0.0156\n",
      "Epoch: 070/100 | Batch 500/600 | Cost: 0.0161\n",
      "Epoch: 070/100 | Batch 550/600 | Cost: 0.0177\n",
      "Epoch: 070/100 Train Cost: 0.0194\n",
      "Time elapsed: 15.98 min\n",
      "Epoch: 071/100 | Batch 000/600 | Cost: 0.0196\n",
      "Epoch: 071/100 | Batch 050/600 | Cost: 0.0232\n",
      "Epoch: 071/100 | Batch 100/600 | Cost: 0.0166\n",
      "Epoch: 071/100 | Batch 150/600 | Cost: 0.0164\n",
      "Epoch: 071/100 | Batch 200/600 | Cost: 0.0177\n",
      "Epoch: 071/100 | Batch 250/600 | Cost: 0.0216\n",
      "Epoch: 071/100 | Batch 300/600 | Cost: 0.0212\n",
      "Epoch: 071/100 | Batch 350/600 | Cost: 0.0168\n",
      "Epoch: 071/100 | Batch 400/600 | Cost: 0.0206\n",
      "Epoch: 071/100 | Batch 450/600 | Cost: 0.0219\n",
      "Epoch: 071/100 | Batch 500/600 | Cost: 0.0200\n",
      "Epoch: 071/100 | Batch 550/600 | Cost: 0.0208\n",
      "Epoch: 071/100 Train Cost: 0.0193\n",
      "Time elapsed: 16.20 min\n",
      "Epoch: 072/100 | Batch 000/600 | Cost: 0.0185\n",
      "Epoch: 072/100 | Batch 050/600 | Cost: 0.0185\n",
      "Epoch: 072/100 | Batch 100/600 | Cost: 0.0252\n",
      "Epoch: 072/100 | Batch 150/600 | Cost: 0.0236\n",
      "Epoch: 072/100 | Batch 200/600 | Cost: 0.0163\n",
      "Epoch: 072/100 | Batch 250/600 | Cost: 0.0179\n",
      "Epoch: 072/100 | Batch 300/600 | Cost: 0.0201\n",
      "Epoch: 072/100 | Batch 350/600 | Cost: 0.0214\n",
      "Epoch: 072/100 | Batch 400/600 | Cost: 0.0134\n",
      "Epoch: 072/100 | Batch 450/600 | Cost: 0.0210\n",
      "Epoch: 072/100 | Batch 500/600 | Cost: 0.0183\n",
      "Epoch: 072/100 | Batch 550/600 | Cost: 0.0182\n",
      "Epoch: 072/100 Train Cost: 0.0192\n",
      "Time elapsed: 16.43 min\n",
      "Epoch: 073/100 | Batch 000/600 | Cost: 0.0183\n",
      "Epoch: 073/100 | Batch 050/600 | Cost: 0.0166\n",
      "Epoch: 073/100 | Batch 100/600 | Cost: 0.0180\n",
      "Epoch: 073/100 | Batch 150/600 | Cost: 0.0197\n",
      "Epoch: 073/100 | Batch 200/600 | Cost: 0.0239\n",
      "Epoch: 073/100 | Batch 250/600 | Cost: 0.0190\n",
      "Epoch: 073/100 | Batch 300/600 | Cost: 0.0139\n",
      "Epoch: 073/100 | Batch 350/600 | Cost: 0.0167\n",
      "Epoch: 073/100 | Batch 400/600 | Cost: 0.0164\n",
      "Epoch: 073/100 | Batch 450/600 | Cost: 0.0223\n",
      "Epoch: 073/100 | Batch 500/600 | Cost: 0.0184\n",
      "Epoch: 073/100 | Batch 550/600 | Cost: 0.0143\n",
      "Epoch: 073/100 Train Cost: 0.0191\n",
      "Time elapsed: 16.66 min\n",
      "Epoch: 074/100 | Batch 000/600 | Cost: 0.0178\n",
      "Epoch: 074/100 | Batch 050/600 | Cost: 0.0189\n",
      "Epoch: 074/100 | Batch 100/600 | Cost: 0.0145\n",
      "Epoch: 074/100 | Batch 150/600 | Cost: 0.0190\n",
      "Epoch: 074/100 | Batch 200/600 | Cost: 0.0233\n",
      "Epoch: 074/100 | Batch 250/600 | Cost: 0.0229\n",
      "Epoch: 074/100 | Batch 300/600 | Cost: 0.0166\n",
      "Epoch: 074/100 | Batch 350/600 | Cost: 0.0193\n",
      "Epoch: 074/100 | Batch 400/600 | Cost: 0.0233\n",
      "Epoch: 074/100 | Batch 450/600 | Cost: 0.0159\n",
      "Epoch: 074/100 | Batch 500/600 | Cost: 0.0157\n",
      "Epoch: 074/100 | Batch 550/600 | Cost: 0.0169\n",
      "Epoch: 074/100 Train Cost: 0.0190\n",
      "Time elapsed: 16.89 min\n",
      "Epoch: 075/100 | Batch 000/600 | Cost: 0.0184\n",
      "Epoch: 075/100 | Batch 050/600 | Cost: 0.0187\n",
      "Epoch: 075/100 | Batch 100/600 | Cost: 0.0221\n",
      "Epoch: 075/100 | Batch 150/600 | Cost: 0.0187\n",
      "Epoch: 075/100 | Batch 200/600 | Cost: 0.0165\n",
      "Epoch: 075/100 | Batch 250/600 | Cost: 0.0124\n",
      "Epoch: 075/100 | Batch 300/600 | Cost: 0.0263\n",
      "Epoch: 075/100 | Batch 350/600 | Cost: 0.0253\n",
      "Epoch: 075/100 | Batch 400/600 | Cost: 0.0163\n",
      "Epoch: 075/100 | Batch 450/600 | Cost: 0.0212\n",
      "Epoch: 075/100 | Batch 500/600 | Cost: 0.0181\n",
      "Epoch: 075/100 | Batch 550/600 | Cost: 0.0129\n",
      "Epoch: 075/100 Train Cost: 0.0189\n",
      "Time elapsed: 17.12 min\n",
      "Epoch: 076/100 | Batch 000/600 | Cost: 0.0177\n",
      "Epoch: 076/100 | Batch 050/600 | Cost: 0.0269\n",
      "Epoch: 076/100 | Batch 100/600 | Cost: 0.0247\n",
      "Epoch: 076/100 | Batch 150/600 | Cost: 0.0182\n",
      "Epoch: 076/100 | Batch 200/600 | Cost: 0.0219\n",
      "Epoch: 076/100 | Batch 250/600 | Cost: 0.0201\n",
      "Epoch: 076/100 | Batch 300/600 | Cost: 0.0139\n",
      "Epoch: 076/100 | Batch 350/600 | Cost: 0.0150\n",
      "Epoch: 076/100 | Batch 400/600 | Cost: 0.0224\n",
      "Epoch: 076/100 | Batch 450/600 | Cost: 0.0144\n",
      "Epoch: 076/100 | Batch 500/600 | Cost: 0.0195\n",
      "Epoch: 076/100 | Batch 550/600 | Cost: 0.0143\n",
      "Epoch: 076/100 Train Cost: 0.0188\n",
      "Time elapsed: 17.35 min\n",
      "Epoch: 077/100 | Batch 000/600 | Cost: 0.0172\n",
      "Epoch: 077/100 | Batch 050/600 | Cost: 0.0230\n",
      "Epoch: 077/100 | Batch 100/600 | Cost: 0.0160\n",
      "Epoch: 077/100 | Batch 150/600 | Cost: 0.0161\n",
      "Epoch: 077/100 | Batch 200/600 | Cost: 0.0238\n",
      "Epoch: 077/100 | Batch 250/600 | Cost: 0.0276\n",
      "Epoch: 077/100 | Batch 300/600 | Cost: 0.0306\n",
      "Epoch: 077/100 | Batch 350/600 | Cost: 0.0173\n",
      "Epoch: 077/100 | Batch 400/600 | Cost: 0.0227\n",
      "Epoch: 077/100 | Batch 450/600 | Cost: 0.0175\n",
      "Epoch: 077/100 | Batch 500/600 | Cost: 0.0132\n",
      "Epoch: 077/100 | Batch 550/600 | Cost: 0.0181\n",
      "Epoch: 077/100 Train Cost: 0.0188\n",
      "Time elapsed: 17.57 min\n",
      "Epoch: 078/100 | Batch 000/600 | Cost: 0.0138\n",
      "Epoch: 078/100 | Batch 050/600 | Cost: 0.0213\n",
      "Epoch: 078/100 | Batch 100/600 | Cost: 0.0178\n",
      "Epoch: 078/100 | Batch 150/600 | Cost: 0.0189\n",
      "Epoch: 078/100 | Batch 200/600 | Cost: 0.0189\n",
      "Epoch: 078/100 | Batch 250/600 | Cost: 0.0169\n",
      "Epoch: 078/100 | Batch 300/600 | Cost: 0.0209\n",
      "Epoch: 078/100 | Batch 350/600 | Cost: 0.0172\n",
      "Epoch: 078/100 | Batch 400/600 | Cost: 0.0194\n",
      "Epoch: 078/100 | Batch 450/600 | Cost: 0.0120\n",
      "Epoch: 078/100 | Batch 500/600 | Cost: 0.0139\n",
      "Epoch: 078/100 | Batch 550/600 | Cost: 0.0250\n",
      "Epoch: 078/100 Train Cost: 0.0187\n",
      "Time elapsed: 17.80 min\n",
      "Epoch: 079/100 | Batch 000/600 | Cost: 0.0216\n",
      "Epoch: 079/100 | Batch 050/600 | Cost: 0.0170\n",
      "Epoch: 079/100 | Batch 100/600 | Cost: 0.0151\n",
      "Epoch: 079/100 | Batch 150/600 | Cost: 0.0196\n",
      "Epoch: 079/100 | Batch 200/600 | Cost: 0.0231\n",
      "Epoch: 079/100 | Batch 250/600 | Cost: 0.0191\n",
      "Epoch: 079/100 | Batch 300/600 | Cost: 0.0208\n",
      "Epoch: 079/100 | Batch 350/600 | Cost: 0.0158\n",
      "Epoch: 079/100 | Batch 400/600 | Cost: 0.0237\n",
      "Epoch: 079/100 | Batch 450/600 | Cost: 0.0137\n",
      "Epoch: 079/100 | Batch 500/600 | Cost: 0.0218\n",
      "Epoch: 079/100 | Batch 550/600 | Cost: 0.0140\n",
      "Epoch: 079/100 Train Cost: 0.0186\n",
      "Time elapsed: 18.03 min\n",
      "Epoch: 080/100 | Batch 000/600 | Cost: 0.0183\n",
      "Epoch: 080/100 | Batch 050/600 | Cost: 0.0165\n",
      "Epoch: 080/100 | Batch 100/600 | Cost: 0.0123\n",
      "Epoch: 080/100 | Batch 150/600 | Cost: 0.0158\n",
      "Epoch: 080/100 | Batch 200/600 | Cost: 0.0154\n",
      "Epoch: 080/100 | Batch 250/600 | Cost: 0.0158\n",
      "Epoch: 080/100 | Batch 300/600 | Cost: 0.0168\n",
      "Epoch: 080/100 | Batch 350/600 | Cost: 0.0139\n",
      "Epoch: 080/100 | Batch 400/600 | Cost: 0.0221\n",
      "Epoch: 080/100 | Batch 450/600 | Cost: 0.0175\n",
      "Epoch: 080/100 | Batch 500/600 | Cost: 0.0151\n",
      "Epoch: 080/100 | Batch 550/600 | Cost: 0.0153\n",
      "Epoch: 080/100 Train Cost: 0.0185\n",
      "Time elapsed: 18.26 min\n",
      "Epoch: 081/100 | Batch 000/600 | Cost: 0.0147\n",
      "Epoch: 081/100 | Batch 050/600 | Cost: 0.0193\n",
      "Epoch: 081/100 | Batch 100/600 | Cost: 0.0100\n",
      "Epoch: 081/100 | Batch 150/600 | Cost: 0.0160\n",
      "Epoch: 081/100 | Batch 200/600 | Cost: 0.0231\n",
      "Epoch: 081/100 | Batch 250/600 | Cost: 0.0129\n",
      "Epoch: 081/100 | Batch 300/600 | Cost: 0.0191\n",
      "Epoch: 081/100 | Batch 350/600 | Cost: 0.0211\n",
      "Epoch: 081/100 | Batch 400/600 | Cost: 0.0183\n",
      "Epoch: 081/100 | Batch 450/600 | Cost: 0.0196\n",
      "Epoch: 081/100 | Batch 500/600 | Cost: 0.0136\n",
      "Epoch: 081/100 | Batch 550/600 | Cost: 0.0138\n",
      "Epoch: 081/100 Train Cost: 0.0184\n",
      "Time elapsed: 18.49 min\n",
      "Epoch: 082/100 | Batch 000/600 | Cost: 0.0140\n",
      "Epoch: 082/100 | Batch 050/600 | Cost: 0.0205\n",
      "Epoch: 082/100 | Batch 100/600 | Cost: 0.0200\n",
      "Epoch: 082/100 | Batch 150/600 | Cost: 0.0164\n",
      "Epoch: 082/100 | Batch 200/600 | Cost: 0.0192\n",
      "Epoch: 082/100 | Batch 250/600 | Cost: 0.0209\n",
      "Epoch: 082/100 | Batch 300/600 | Cost: 0.0173\n",
      "Epoch: 082/100 | Batch 350/600 | Cost: 0.0200\n",
      "Epoch: 082/100 | Batch 400/600 | Cost: 0.0154\n",
      "Epoch: 082/100 | Batch 450/600 | Cost: 0.0257\n",
      "Epoch: 082/100 | Batch 500/600 | Cost: 0.0151\n",
      "Epoch: 082/100 | Batch 550/600 | Cost: 0.0176\n",
      "Epoch: 082/100 Train Cost: 0.0183\n",
      "Time elapsed: 18.72 min\n",
      "Epoch: 083/100 | Batch 000/600 | Cost: 0.0160\n",
      "Epoch: 083/100 | Batch 050/600 | Cost: 0.0202\n",
      "Epoch: 083/100 | Batch 100/600 | Cost: 0.0119\n",
      "Epoch: 083/100 | Batch 150/600 | Cost: 0.0179\n",
      "Epoch: 083/100 | Batch 200/600 | Cost: 0.0167\n",
      "Epoch: 083/100 | Batch 250/600 | Cost: 0.0166\n",
      "Epoch: 083/100 | Batch 300/600 | Cost: 0.0214\n",
      "Epoch: 083/100 | Batch 350/600 | Cost: 0.0138\n",
      "Epoch: 083/100 | Batch 400/600 | Cost: 0.0160\n",
      "Epoch: 083/100 | Batch 450/600 | Cost: 0.0202\n",
      "Epoch: 083/100 | Batch 500/600 | Cost: 0.0161\n",
      "Epoch: 083/100 | Batch 550/600 | Cost: 0.0182\n",
      "Epoch: 083/100 Train Cost: 0.0183\n",
      "Time elapsed: 18.94 min\n",
      "Epoch: 084/100 | Batch 000/600 | Cost: 0.0200\n",
      "Epoch: 084/100 | Batch 050/600 | Cost: 0.0115\n",
      "Epoch: 084/100 | Batch 100/600 | Cost: 0.0182\n",
      "Epoch: 084/100 | Batch 150/600 | Cost: 0.0147\n",
      "Epoch: 084/100 | Batch 200/600 | Cost: 0.0171\n",
      "Epoch: 084/100 | Batch 250/600 | Cost: 0.0188\n",
      "Epoch: 084/100 | Batch 300/600 | Cost: 0.0183\n",
      "Epoch: 084/100 | Batch 350/600 | Cost: 0.0187\n",
      "Epoch: 084/100 | Batch 400/600 | Cost: 0.0175\n",
      "Epoch: 084/100 | Batch 450/600 | Cost: 0.0186\n",
      "Epoch: 084/100 | Batch 500/600 | Cost: 0.0230\n",
      "Epoch: 084/100 | Batch 550/600 | Cost: 0.0156\n",
      "Epoch: 084/100 Train Cost: 0.0182\n",
      "Time elapsed: 19.17 min\n",
      "Epoch: 085/100 | Batch 000/600 | Cost: 0.0165\n",
      "Epoch: 085/100 | Batch 050/600 | Cost: 0.0213\n",
      "Epoch: 085/100 | Batch 100/600 | Cost: 0.0173\n",
      "Epoch: 085/100 | Batch 150/600 | Cost: 0.0162\n",
      "Epoch: 085/100 | Batch 200/600 | Cost: 0.0161\n",
      "Epoch: 085/100 | Batch 250/600 | Cost: 0.0188\n",
      "Epoch: 085/100 | Batch 300/600 | Cost: 0.0198\n",
      "Epoch: 085/100 | Batch 350/600 | Cost: 0.0155\n",
      "Epoch: 085/100 | Batch 400/600 | Cost: 0.0194\n",
      "Epoch: 085/100 | Batch 450/600 | Cost: 0.0178\n",
      "Epoch: 085/100 | Batch 500/600 | Cost: 0.0191\n",
      "Epoch: 085/100 | Batch 550/600 | Cost: 0.0226\n",
      "Epoch: 085/100 Train Cost: 0.0181\n",
      "Time elapsed: 19.40 min\n",
      "Epoch: 086/100 | Batch 000/600 | Cost: 0.0209\n",
      "Epoch: 086/100 | Batch 050/600 | Cost: 0.0199\n",
      "Epoch: 086/100 | Batch 100/600 | Cost: 0.0132\n",
      "Epoch: 086/100 | Batch 150/600 | Cost: 0.0144\n",
      "Epoch: 086/100 | Batch 200/600 | Cost: 0.0137\n",
      "Epoch: 086/100 | Batch 250/600 | Cost: 0.0195\n",
      "Epoch: 086/100 | Batch 300/600 | Cost: 0.0248\n",
      "Epoch: 086/100 | Batch 350/600 | Cost: 0.0181\n",
      "Epoch: 086/100 | Batch 400/600 | Cost: 0.0181\n",
      "Epoch: 086/100 | Batch 450/600 | Cost: 0.0187\n",
      "Epoch: 086/100 | Batch 500/600 | Cost: 0.0173\n",
      "Epoch: 086/100 | Batch 550/600 | Cost: 0.0161\n",
      "Epoch: 086/100 Train Cost: 0.0180\n",
      "Time elapsed: 19.63 min\n",
      "Epoch: 087/100 | Batch 000/600 | Cost: 0.0182\n",
      "Epoch: 087/100 | Batch 050/600 | Cost: 0.0176\n",
      "Epoch: 087/100 | Batch 100/600 | Cost: 0.0155\n",
      "Epoch: 087/100 | Batch 150/600 | Cost: 0.0173\n",
      "Epoch: 087/100 | Batch 200/600 | Cost: 0.0196\n",
      "Epoch: 087/100 | Batch 250/600 | Cost: 0.0146\n",
      "Epoch: 087/100 | Batch 300/600 | Cost: 0.0194\n",
      "Epoch: 087/100 | Batch 350/600 | Cost: 0.0137\n",
      "Epoch: 087/100 | Batch 400/600 | Cost: 0.0195\n",
      "Epoch: 087/100 | Batch 450/600 | Cost: 0.0146\n",
      "Epoch: 087/100 | Batch 500/600 | Cost: 0.0143\n",
      "Epoch: 087/100 | Batch 550/600 | Cost: 0.0129\n",
      "Epoch: 087/100 Train Cost: 0.0180\n",
      "Time elapsed: 19.87 min\n",
      "Epoch: 088/100 | Batch 000/600 | Cost: 0.0104\n",
      "Epoch: 088/100 | Batch 050/600 | Cost: 0.0121\n",
      "Epoch: 088/100 | Batch 100/600 | Cost: 0.0179\n",
      "Epoch: 088/100 | Batch 150/600 | Cost: 0.0206\n",
      "Epoch: 088/100 | Batch 200/600 | Cost: 0.0155\n",
      "Epoch: 088/100 | Batch 250/600 | Cost: 0.0132\n",
      "Epoch: 088/100 | Batch 300/600 | Cost: 0.0150\n",
      "Epoch: 088/100 | Batch 350/600 | Cost: 0.0196\n",
      "Epoch: 088/100 | Batch 400/600 | Cost: 0.0137\n",
      "Epoch: 088/100 | Batch 450/600 | Cost: 0.0170\n",
      "Epoch: 088/100 | Batch 500/600 | Cost: 0.0176\n",
      "Epoch: 088/100 | Batch 550/600 | Cost: 0.0213\n",
      "Epoch: 088/100 Train Cost: 0.0179\n",
      "Time elapsed: 20.10 min\n",
      "Epoch: 089/100 | Batch 000/600 | Cost: 0.0164\n",
      "Epoch: 089/100 | Batch 050/600 | Cost: 0.0235\n",
      "Epoch: 089/100 | Batch 100/600 | Cost: 0.0196\n",
      "Epoch: 089/100 | Batch 150/600 | Cost: 0.0171\n",
      "Epoch: 089/100 | Batch 200/600 | Cost: 0.0212\n",
      "Epoch: 089/100 | Batch 250/600 | Cost: 0.0173\n",
      "Epoch: 089/100 | Batch 300/600 | Cost: 0.0165\n",
      "Epoch: 089/100 | Batch 350/600 | Cost: 0.0236\n",
      "Epoch: 089/100 | Batch 400/600 | Cost: 0.0197\n",
      "Epoch: 089/100 | Batch 450/600 | Cost: 0.0249\n",
      "Epoch: 089/100 | Batch 500/600 | Cost: 0.0147\n",
      "Epoch: 089/100 | Batch 550/600 | Cost: 0.0192\n",
      "Epoch: 089/100 Train Cost: 0.0178\n",
      "Time elapsed: 20.33 min\n",
      "Epoch: 090/100 | Batch 000/600 | Cost: 0.0206\n",
      "Epoch: 090/100 | Batch 050/600 | Cost: 0.0176\n",
      "Epoch: 090/100 | Batch 100/600 | Cost: 0.0154\n",
      "Epoch: 090/100 | Batch 150/600 | Cost: 0.0164\n",
      "Epoch: 090/100 | Batch 200/600 | Cost: 0.0136\n",
      "Epoch: 090/100 | Batch 250/600 | Cost: 0.0155\n",
      "Epoch: 090/100 | Batch 300/600 | Cost: 0.0152\n",
      "Epoch: 090/100 | Batch 350/600 | Cost: 0.0191\n",
      "Epoch: 090/100 | Batch 400/600 | Cost: 0.0190\n",
      "Epoch: 090/100 | Batch 450/600 | Cost: 0.0157\n",
      "Epoch: 090/100 | Batch 500/600 | Cost: 0.0107\n",
      "Epoch: 090/100 | Batch 550/600 | Cost: 0.0186\n",
      "Epoch: 090/100 Train Cost: 0.0178\n",
      "Time elapsed: 20.55 min\n",
      "Epoch: 091/100 | Batch 000/600 | Cost: 0.0151\n",
      "Epoch: 091/100 | Batch 050/600 | Cost: 0.0175\n",
      "Epoch: 091/100 | Batch 100/600 | Cost: 0.0180\n",
      "Epoch: 091/100 | Batch 150/600 | Cost: 0.0218\n",
      "Epoch: 091/100 | Batch 200/600 | Cost: 0.0186\n",
      "Epoch: 091/100 | Batch 250/600 | Cost: 0.0147\n",
      "Epoch: 091/100 | Batch 300/600 | Cost: 0.0142\n",
      "Epoch: 091/100 | Batch 350/600 | Cost: 0.0208\n",
      "Epoch: 091/100 | Batch 400/600 | Cost: 0.0223\n",
      "Epoch: 091/100 | Batch 450/600 | Cost: 0.0172\n",
      "Epoch: 091/100 | Batch 500/600 | Cost: 0.0193\n",
      "Epoch: 091/100 | Batch 550/600 | Cost: 0.0171\n",
      "Epoch: 091/100 Train Cost: 0.0177\n",
      "Time elapsed: 20.78 min\n",
      "Epoch: 092/100 | Batch 000/600 | Cost: 0.0197\n",
      "Epoch: 092/100 | Batch 050/600 | Cost: 0.0177\n",
      "Epoch: 092/100 | Batch 100/600 | Cost: 0.0202\n",
      "Epoch: 092/100 | Batch 150/600 | Cost: 0.0172\n",
      "Epoch: 092/100 | Batch 200/600 | Cost: 0.0161\n",
      "Epoch: 092/100 | Batch 250/600 | Cost: 0.0175\n",
      "Epoch: 092/100 | Batch 300/600 | Cost: 0.0134\n",
      "Epoch: 092/100 | Batch 350/600 | Cost: 0.0175\n",
      "Epoch: 092/100 | Batch 400/600 | Cost: 0.0204\n",
      "Epoch: 092/100 | Batch 450/600 | Cost: 0.0184\n",
      "Epoch: 092/100 | Batch 500/600 | Cost: 0.0177\n",
      "Epoch: 092/100 | Batch 550/600 | Cost: 0.0207\n",
      "Epoch: 092/100 Train Cost: 0.0176\n",
      "Time elapsed: 21.01 min\n",
      "Epoch: 093/100 | Batch 000/600 | Cost: 0.0189\n",
      "Epoch: 093/100 | Batch 050/600 | Cost: 0.0230\n",
      "Epoch: 093/100 | Batch 100/600 | Cost: 0.0156\n",
      "Epoch: 093/100 | Batch 150/600 | Cost: 0.0193\n",
      "Epoch: 093/100 | Batch 200/600 | Cost: 0.0181\n",
      "Epoch: 093/100 | Batch 250/600 | Cost: 0.0145\n",
      "Epoch: 093/100 | Batch 300/600 | Cost: 0.0238\n",
      "Epoch: 093/100 | Batch 350/600 | Cost: 0.0219\n",
      "Epoch: 093/100 | Batch 400/600 | Cost: 0.0188\n",
      "Epoch: 093/100 | Batch 450/600 | Cost: 0.0171\n",
      "Epoch: 093/100 | Batch 500/600 | Cost: 0.0208\n",
      "Epoch: 093/100 | Batch 550/600 | Cost: 0.0181\n",
      "Epoch: 093/100 Train Cost: 0.0176\n",
      "Time elapsed: 21.24 min\n",
      "Epoch: 094/100 | Batch 000/600 | Cost: 0.0201\n",
      "Epoch: 094/100 | Batch 050/600 | Cost: 0.0144\n",
      "Epoch: 094/100 | Batch 100/600 | Cost: 0.0182\n",
      "Epoch: 094/100 | Batch 150/600 | Cost: 0.0195\n",
      "Epoch: 094/100 | Batch 200/600 | Cost: 0.0202\n",
      "Epoch: 094/100 | Batch 250/600 | Cost: 0.0224\n",
      "Epoch: 094/100 | Batch 300/600 | Cost: 0.0153\n",
      "Epoch: 094/100 | Batch 350/600 | Cost: 0.0161\n",
      "Epoch: 094/100 | Batch 400/600 | Cost: 0.0162\n",
      "Epoch: 094/100 | Batch 450/600 | Cost: 0.0189\n",
      "Epoch: 094/100 | Batch 500/600 | Cost: 0.0178\n",
      "Epoch: 094/100 | Batch 550/600 | Cost: 0.0151\n",
      "Epoch: 094/100 Train Cost: 0.0175\n",
      "Time elapsed: 21.47 min\n",
      "Epoch: 095/100 | Batch 000/600 | Cost: 0.0167\n",
      "Epoch: 095/100 | Batch 050/600 | Cost: 0.0148\n",
      "Epoch: 095/100 | Batch 100/600 | Cost: 0.0138\n",
      "Epoch: 095/100 | Batch 150/600 | Cost: 0.0134\n",
      "Epoch: 095/100 | Batch 200/600 | Cost: 0.0159\n",
      "Epoch: 095/100 | Batch 250/600 | Cost: 0.0138\n",
      "Epoch: 095/100 | Batch 300/600 | Cost: 0.0149\n",
      "Epoch: 095/100 | Batch 350/600 | Cost: 0.0204\n",
      "Epoch: 095/100 | Batch 400/600 | Cost: 0.0178\n",
      "Epoch: 095/100 | Batch 450/600 | Cost: 0.0127\n",
      "Epoch: 095/100 | Batch 500/600 | Cost: 0.0183\n",
      "Epoch: 095/100 | Batch 550/600 | Cost: 0.0188\n",
      "Epoch: 095/100 Train Cost: 0.0174\n",
      "Time elapsed: 21.69 min\n",
      "Epoch: 096/100 | Batch 000/600 | Cost: 0.0187\n",
      "Epoch: 096/100 | Batch 050/600 | Cost: 0.0225\n",
      "Epoch: 096/100 | Batch 100/600 | Cost: 0.0128\n",
      "Epoch: 096/100 | Batch 150/600 | Cost: 0.0162\n",
      "Epoch: 096/100 | Batch 200/600 | Cost: 0.0232\n",
      "Epoch: 096/100 | Batch 250/600 | Cost: 0.0147\n",
      "Epoch: 096/100 | Batch 300/600 | Cost: 0.0169\n",
      "Epoch: 096/100 | Batch 350/600 | Cost: 0.0154\n",
      "Epoch: 096/100 | Batch 400/600 | Cost: 0.0217\n",
      "Epoch: 096/100 | Batch 450/600 | Cost: 0.0155\n",
      "Epoch: 096/100 | Batch 500/600 | Cost: 0.0160\n",
      "Epoch: 096/100 | Batch 550/600 | Cost: 0.0185\n",
      "Epoch: 096/100 Train Cost: 0.0174\n",
      "Time elapsed: 21.92 min\n",
      "Epoch: 097/100 | Batch 000/600 | Cost: 0.0168\n",
      "Epoch: 097/100 | Batch 050/600 | Cost: 0.0170\n",
      "Epoch: 097/100 | Batch 100/600 | Cost: 0.0214\n",
      "Epoch: 097/100 | Batch 150/600 | Cost: 0.0183\n",
      "Epoch: 097/100 | Batch 200/600 | Cost: 0.0161\n",
      "Epoch: 097/100 | Batch 250/600 | Cost: 0.0197\n",
      "Epoch: 097/100 | Batch 300/600 | Cost: 0.0157\n",
      "Epoch: 097/100 | Batch 350/600 | Cost: 0.0216\n",
      "Epoch: 097/100 | Batch 400/600 | Cost: 0.0161\n",
      "Epoch: 097/100 | Batch 450/600 | Cost: 0.0146\n",
      "Epoch: 097/100 | Batch 500/600 | Cost: 0.0140\n",
      "Epoch: 097/100 | Batch 550/600 | Cost: 0.0215\n",
      "Epoch: 097/100 Train Cost: 0.0173\n",
      "Time elapsed: 22.16 min\n",
      "Epoch: 098/100 | Batch 000/600 | Cost: 0.0208\n",
      "Epoch: 098/100 | Batch 050/600 | Cost: 0.0151\n",
      "Epoch: 098/100 | Batch 100/600 | Cost: 0.0218\n",
      "Epoch: 098/100 | Batch 150/600 | Cost: 0.0228\n",
      "Epoch: 098/100 | Batch 200/600 | Cost: 0.0164\n",
      "Epoch: 098/100 | Batch 250/600 | Cost: 0.0147\n",
      "Epoch: 098/100 | Batch 300/600 | Cost: 0.0162\n",
      "Epoch: 098/100 | Batch 350/600 | Cost: 0.0255\n",
      "Epoch: 098/100 | Batch 400/600 | Cost: 0.0202\n",
      "Epoch: 098/100 | Batch 450/600 | Cost: 0.0127\n",
      "Epoch: 098/100 | Batch 500/600 | Cost: 0.0167\n",
      "Epoch: 098/100 | Batch 550/600 | Cost: 0.0218\n",
      "Epoch: 098/100 Train Cost: 0.0173\n",
      "Time elapsed: 22.39 min\n",
      "Epoch: 099/100 | Batch 000/600 | Cost: 0.0168\n",
      "Epoch: 099/100 | Batch 050/600 | Cost: 0.0205\n",
      "Epoch: 099/100 | Batch 100/600 | Cost: 0.0134\n",
      "Epoch: 099/100 | Batch 150/600 | Cost: 0.0181\n",
      "Epoch: 099/100 | Batch 200/600 | Cost: 0.0162\n",
      "Epoch: 099/100 | Batch 250/600 | Cost: 0.0194\n",
      "Epoch: 099/100 | Batch 300/600 | Cost: 0.0167\n",
      "Epoch: 099/100 | Batch 350/600 | Cost: 0.0165\n",
      "Epoch: 099/100 | Batch 400/600 | Cost: 0.0226\n",
      "Epoch: 099/100 | Batch 450/600 | Cost: 0.0155\n",
      "Epoch: 099/100 | Batch 500/600 | Cost: 0.0143\n",
      "Epoch: 099/100 | Batch 550/600 | Cost: 0.0188\n",
      "Epoch: 099/100 Train Cost: 0.0172\n",
      "Time elapsed: 22.62 min\n",
      "Epoch: 100/100 | Batch 000/600 | Cost: 0.0159\n",
      "Epoch: 100/100 | Batch 050/600 | Cost: 0.0188\n",
      "Epoch: 100/100 | Batch 100/600 | Cost: 0.0144\n",
      "Epoch: 100/100 | Batch 150/600 | Cost: 0.0166\n",
      "Epoch: 100/100 | Batch 200/600 | Cost: 0.0209\n",
      "Epoch: 100/100 | Batch 250/600 | Cost: 0.0202\n",
      "Epoch: 100/100 | Batch 300/600 | Cost: 0.0177\n",
      "Epoch: 100/100 | Batch 350/600 | Cost: 0.0194\n",
      "Epoch: 100/100 | Batch 400/600 | Cost: 0.0173\n",
      "Epoch: 100/100 | Batch 450/600 | Cost: 0.0194\n",
      "Epoch: 100/100 | Batch 500/600 | Cost: 0.0193\n",
      "Epoch: 100/100 | Batch 550/600 | Cost: 0.0124\n",
      "Epoch: 100/100 Train Cost: 0.0171\n",
      "Time elapsed: 22.85 min\n",
      "Total Training Time: 22.85 min\n"
     ]
    }
   ],
   "source": [
    "class MlpSigmoidMSE(torch.nn.Module):\n",
    "\n",
    "    def __init__(self, num_features, num_hidden, num_classes):\n",
    "        super(MlpSigmoidMSE, self).__init__()\n",
    "        \n",
    "        self.num_classes = num_classes\n",
    "        \n",
    "        ### 1st hidden layer\n",
    "        self.linear_1 = torch.nn.Linear(num_features, num_hidden)\n",
    "        self.linear_1.weight.detach().normal_(0.0, 0.1)\n",
    "        self.linear_1.bias.detach().zero_()\n",
    "\n",
    "        ### Output layer\n",
    "        self.linear_out = torch.nn.Linear(num_hidden, num_classes)\n",
    "        self.linear_out.weight.detach().normal_(0.0, 0.1)\n",
    "        self.linear_out.bias.detach().zero_()\n",
    "        \n",
    "    def forward(self, x):\n",
    "        out = self.linear_1(x)\n",
    "        out = torch.sigmoid(out)\n",
    "        logits = self.linear_out(out)\n",
    "        probas = torch.sigmoid(logits)\n",
    "        return logits, probas\n",
    "\n",
    "    \n",
    "#################################\n",
    "### Model Initialization\n",
    "#################################\n",
    "    \n",
    "torch.manual_seed(RANDOM_SEED)\n",
    "model = MlpSigmoidMSE(num_features=28*28,\n",
    "                      num_hidden=100,\n",
    "                      num_classes=10)\n",
    "\n",
    "model = model.to(DEVICE)\n",
    "\n",
    "optimizer = torch.optim.SGD(model.parameters(), lr=0.1)\n",
    "\n",
    "#################################\n",
    "### Training\n",
    "#################################\n",
    "\n",
    "def compute_mse(net, data_loader):\n",
    "    curr_mse, num_examples = torch.zeros(model.num_classes).float(), 0\n",
    "    with torch.no_grad():\n",
    "        for features, targets in data_loader:\n",
    "            features = features.view(-1, 28*28).to(DEVICE)\n",
    "            logits, probas = net.forward(features)\n",
    "            probas = probas.to(torch.device('cpu'))\n",
    "            y_onehot = to_onehot(targets, model.num_classes)\n",
    "            loss = torch.sum((y_onehot - probas)**2, dim=0)\n",
    "            num_examples += targets.size(0)\n",
    "            curr_mse += loss\n",
    "\n",
    "        curr_mse = torch.mean(curr_mse/num_examples, dim=0)\n",
    "        return curr_mse\n",
    "\n",
    "\n",
    "\n",
    "start_time = time.time()\n",
    "minibatch_cost = []\n",
    "epoch_cost = []\n",
    "for epoch in range(NUM_EPOCHS):\n",
    "    model.train()\n",
    "    for batch_idx, (features, targets) in enumerate(train_loader):\n",
    "        \n",
    "        features = features.view(-1, 28*28).to(DEVICE)\n",
    "            \n",
    "        ### FORWARD AND BACK PROP\n",
    "        logits, probas = model(features)\n",
    "        \n",
    "        y_onehot = to_onehot(targets, model.num_classes).to(DEVICE)\n",
    "        \n",
    "        cost = F.mse_loss(y_onehot, probas)\n",
    "        optimizer.zero_grad()\n",
    "        \n",
    "        cost.backward()\n",
    "        minibatch_cost.append(cost)\n",
    "        ### UPDATE MODEL PARAMETERS\n",
    "        optimizer.step()\n",
    "        \n",
    "        ### LOGGING\n",
    "        if not batch_idx % 50:\n",
    "            print ('Epoch: %03d/%03d | Batch %03d/%03d | Cost: %.4f' \n",
    "                   %(epoch+1, NUM_EPOCHS, batch_idx, \n",
    "                     len(train_loader), cost))\n",
    "        \n",
    "    cost = compute_mse(model, train_loader)\n",
    "    epoch_cost.append(cost)\n",
    "    print('Epoch: %03d/%03d Train Cost: %.4f' % (\n",
    "            epoch+1, NUM_EPOCHS, cost))\n",
    "    print('Time elapsed: %.2f min' % ((time.time() - start_time)/60))\n",
    "    \n",
    "print('Total Training Time: %.2f min' % ((time.time() - start_time)/60))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(range(len(minibatch_cost)), minibatch_cost)\n",
    "plt.ylabel('Mean Squared Error')\n",
    "plt.xlabel('Minibatch')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(range(len(epoch_cost)), epoch_cost)\n",
    "plt.ylabel('Mean Squared Error')\n",
    "plt.xlabel('Epoch')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuracy: 90.78\n",
      "Test Accuracy: 91.34\n"
     ]
    }
   ],
   "source": [
    "def compute_accuracy(net, data_loader):\n",
    "    correct_pred, num_examples = 0, 0\n",
    "    with torch.no_grad():\n",
    "        for features, targets in data_loader:\n",
    "            features = features.view(-1, 28*28).to(DEVICE)\n",
    "            targets = targets.to(DEVICE)\n",
    "            a1, a2 = net.forward(features)\n",
    "            predicted_labels = torch.argmax(a2, 1)\n",
    "            num_examples += targets.size(0)\n",
    "            correct_pred += (predicted_labels == targets).sum()\n",
    "        return correct_pred.float()/num_examples * 100\n",
    "    \n",
    "print('Training Accuracy: %.2f' % compute_accuracy(model, train_loader))\n",
    "print('Test Accuracy: %.2f' % compute_accuracy(model, test_loader))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Sigmoid Activation with Cross Entropy Loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(range(len(minibatch_cost)), minibatch_cost)\n",
    "plt.ylabel('Cross Entropy')\n",
    "plt.xlabel('Minibatch')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(range(len(epoch_cost)), epoch_cost)\n",
    "plt.ylabel('Cross Entropy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuracy: 90.78\n",
      "Test Accuracy: 91.34\n"
     ]
    }
   ],
   "source": [
    "def compute_accuracy(net, data_loader):\n",
    "    correct_pred, num_examples = 0, 0\n",
    "    with torch.no_grad():\n",
    "        for features, targets in data_loader:\n",
    "            features = features.view(-1, 28*28).to(DEVICE)\n",
    "            targets = targets.to(DEVICE)\n",
    "            a1, a2 = net.forward(features)\n",
    "            predicted_labels = torch.argmax(a2, 1)\n",
    "            num_examples += targets.size(0)\n",
    "            correct_pred += (predicted_labels == targets).sum()\n",
    "        return correct_pred.float()/num_examples * 100\n",
    "    \n",
    "print('Training Accuracy: %.2f' % compute_accuracy(model, train_loader))\n",
    "print('Test Accuracy: %.2f' % compute_accuracy(model, test_loader))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Sigmoid Activation with Sofmax output and Cross Entropy Loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 001/100 | Batch 000/600 | Cost: 2.3579\n",
      "Epoch: 001/100 | Batch 050/600 | Cost: 1.8853\n",
      "Epoch: 001/100 | Batch 100/600 | Cost: 1.4858\n",
      "Epoch: 001/100 | Batch 150/600 | Cost: 1.0450\n",
      "Epoch: 001/100 | Batch 200/600 | Cost: 0.9816\n",
      "Epoch: 001/100 | Batch 250/600 | Cost: 0.8292\n",
      "Epoch: 001/100 | Batch 300/600 | Cost: 0.7118\n",
      "Epoch: 001/100 | Batch 350/600 | Cost: 0.6683\n",
      "Epoch: 001/100 | Batch 400/600 | Cost: 0.4790\n",
      "Epoch: 001/100 | Batch 450/600 | Cost: 0.6711\n",
      "Epoch: 001/100 | Batch 500/600 | Cost: 0.4821\n",
      "Epoch: 001/100 | Batch 550/600 | Cost: 0.5106\n",
      "Epoch: 001/100 Train Cost: 0.4951\n",
      "Time elapsed: 0.23 min\n",
      "Epoch: 002/100 | Batch 000/600 | Cost: 0.5281\n",
      "Epoch: 002/100 | Batch 050/600 | Cost: 0.3814\n",
      "Epoch: 002/100 | Batch 100/600 | Cost: 0.5348\n",
      "Epoch: 002/100 | Batch 150/600 | Cost: 0.3768\n",
      "Epoch: 002/100 | Batch 200/600 | Cost: 0.3754\n",
      "Epoch: 002/100 | Batch 250/600 | Cost: 0.4553\n",
      "Epoch: 002/100 | Batch 300/600 | Cost: 0.4784\n",
      "Epoch: 002/100 | Batch 350/600 | Cost: 0.2455\n",
      "Epoch: 002/100 | Batch 400/600 | Cost: 0.2256\n",
      "Epoch: 002/100 | Batch 450/600 | Cost: 0.3198\n",
      "Epoch: 002/100 | Batch 500/600 | Cost: 0.4637\n",
      "Epoch: 002/100 | Batch 550/600 | Cost: 0.3965\n",
      "Epoch: 002/100 Train Cost: 0.3739\n",
      "Time elapsed: 0.46 min\n",
      "Epoch: 003/100 | Batch 000/600 | Cost: 0.3823\n",
      "Epoch: 003/100 | Batch 050/600 | Cost: 0.4210\n",
      "Epoch: 003/100 | Batch 100/600 | Cost: 0.3343\n",
      "Epoch: 003/100 | Batch 150/600 | Cost: 0.4236\n",
      "Epoch: 003/100 | Batch 200/600 | Cost: 0.3640\n",
      "Epoch: 003/100 | Batch 250/600 | Cost: 0.2810\n",
      "Epoch: 003/100 | Batch 300/600 | Cost: 0.3518\n",
      "Epoch: 003/100 | Batch 350/600 | Cost: 0.4136\n",
      "Epoch: 003/100 | Batch 400/600 | Cost: 0.3330\n",
      "Epoch: 003/100 | Batch 450/600 | Cost: 0.3568\n",
      "Epoch: 003/100 | Batch 500/600 | Cost: 0.4766\n",
      "Epoch: 003/100 | Batch 550/600 | Cost: 0.4609\n",
      "Epoch: 003/100 Train Cost: 0.3302\n",
      "Time elapsed: 0.69 min\n",
      "Epoch: 004/100 | Batch 000/600 | Cost: 0.2824\n",
      "Epoch: 004/100 | Batch 050/600 | Cost: 0.4340\n",
      "Epoch: 004/100 | Batch 100/600 | Cost: 0.3058\n",
      "Epoch: 004/100 | Batch 150/600 | Cost: 0.3312\n",
      "Epoch: 004/100 | Batch 200/600 | Cost: 0.3155\n",
      "Epoch: 004/100 | Batch 250/600 | Cost: 0.3839\n",
      "Epoch: 004/100 | Batch 300/600 | Cost: 0.2353\n",
      "Epoch: 004/100 | Batch 350/600 | Cost: 0.2168\n",
      "Epoch: 004/100 | Batch 400/600 | Cost: 0.2734\n",
      "Epoch: 004/100 | Batch 450/600 | Cost: 0.3435\n",
      "Epoch: 004/100 | Batch 500/600 | Cost: 0.4454\n",
      "Epoch: 004/100 | Batch 550/600 | Cost: 0.4063\n",
      "Epoch: 004/100 Train Cost: 0.3037\n",
      "Time elapsed: 0.92 min\n",
      "Epoch: 005/100 | Batch 000/600 | Cost: 0.2349\n",
      "Epoch: 005/100 | Batch 050/600 | Cost: 0.2860\n",
      "Epoch: 005/100 | Batch 100/600 | Cost: 0.2677\n",
      "Epoch: 005/100 | Batch 150/600 | Cost: 0.2220\n",
      "Epoch: 005/100 | Batch 200/600 | Cost: 0.1747\n",
      "Epoch: 005/100 | Batch 250/600 | Cost: 0.3912\n",
      "Epoch: 005/100 | Batch 300/600 | Cost: 0.3540\n",
      "Epoch: 005/100 | Batch 350/600 | Cost: 0.2502\n",
      "Epoch: 005/100 | Batch 400/600 | Cost: 0.1948\n",
      "Epoch: 005/100 | Batch 450/600 | Cost: 0.2429\n",
      "Epoch: 005/100 | Batch 500/600 | Cost: 0.3263\n",
      "Epoch: 005/100 | Batch 550/600 | Cost: 0.2342\n",
      "Epoch: 005/100 Train Cost: 0.2860\n",
      "Time elapsed: 1.15 min\n",
      "Epoch: 006/100 | Batch 000/600 | Cost: 0.1479\n",
      "Epoch: 006/100 | Batch 050/600 | Cost: 0.1950\n",
      "Epoch: 006/100 | Batch 100/600 | Cost: 0.1540\n",
      "Epoch: 006/100 | Batch 150/600 | Cost: 0.2886\n",
      "Epoch: 006/100 | Batch 200/600 | Cost: 0.2719\n",
      "Epoch: 006/100 | Batch 250/600 | Cost: 0.3163\n",
      "Epoch: 006/100 | Batch 300/600 | Cost: 0.2726\n",
      "Epoch: 006/100 | Batch 350/600 | Cost: 0.3299\n",
      "Epoch: 006/100 | Batch 400/600 | Cost: 0.2617\n",
      "Epoch: 006/100 | Batch 450/600 | Cost: 0.2420\n",
      "Epoch: 006/100 | Batch 500/600 | Cost: 0.3257\n",
      "Epoch: 006/100 | Batch 550/600 | Cost: 0.2581\n",
      "Epoch: 006/100 Train Cost: 0.2703\n",
      "Time elapsed: 1.37 min\n",
      "Epoch: 007/100 | Batch 000/600 | Cost: 0.2868\n",
      "Epoch: 007/100 | Batch 050/600 | Cost: 0.1424\n",
      "Epoch: 007/100 | Batch 100/600 | Cost: 0.2341\n",
      "Epoch: 007/100 | Batch 150/600 | Cost: 0.3194\n",
      "Epoch: 007/100 | Batch 200/600 | Cost: 0.1871\n",
      "Epoch: 007/100 | Batch 250/600 | Cost: 0.2656\n",
      "Epoch: 007/100 | Batch 300/600 | Cost: 0.1738\n",
      "Epoch: 007/100 | Batch 350/600 | Cost: 0.2957\n",
      "Epoch: 007/100 | Batch 400/600 | Cost: 0.2223\n",
      "Epoch: 007/100 | Batch 450/600 | Cost: 0.2679\n",
      "Epoch: 007/100 | Batch 500/600 | Cost: 0.2304\n",
      "Epoch: 007/100 | Batch 550/600 | Cost: 0.3339\n",
      "Epoch: 007/100 Train Cost: 0.2570\n",
      "Time elapsed: 1.60 min\n",
      "Epoch: 008/100 | Batch 000/600 | Cost: 0.1670\n",
      "Epoch: 008/100 | Batch 050/600 | Cost: 0.2860\n",
      "Epoch: 008/100 | Batch 100/600 | Cost: 0.2400\n",
      "Epoch: 008/100 | Batch 150/600 | Cost: 0.2061\n",
      "Epoch: 008/100 | Batch 200/600 | Cost: 0.3341\n",
      "Epoch: 008/100 | Batch 250/600 | Cost: 0.3889\n",
      "Epoch: 008/100 | Batch 300/600 | Cost: 0.3033\n",
      "Epoch: 008/100 | Batch 350/600 | Cost: 0.2118\n",
      "Epoch: 008/100 | Batch 400/600 | Cost: 0.2464\n",
      "Epoch: 008/100 | Batch 450/600 | Cost: 0.2186\n",
      "Epoch: 008/100 | Batch 500/600 | Cost: 0.1369\n",
      "Epoch: 008/100 | Batch 550/600 | Cost: 0.3552\n",
      "Epoch: 008/100 Train Cost: 0.2453\n",
      "Time elapsed: 1.83 min\n",
      "Epoch: 009/100 | Batch 000/600 | Cost: 0.1667\n",
      "Epoch: 009/100 | Batch 050/600 | Cost: 0.2755\n",
      "Epoch: 009/100 | Batch 100/600 | Cost: 0.2297\n",
      "Epoch: 009/100 | Batch 150/600 | Cost: 0.1814\n",
      "Epoch: 009/100 | Batch 200/600 | Cost: 0.2220\n",
      "Epoch: 009/100 | Batch 250/600 | Cost: 0.2005\n",
      "Epoch: 009/100 | Batch 300/600 | Cost: 0.2196\n",
      "Epoch: 009/100 | Batch 350/600 | Cost: 0.2688\n",
      "Epoch: 009/100 | Batch 400/600 | Cost: 0.4307\n",
      "Epoch: 009/100 | Batch 450/600 | Cost: 0.2187\n",
      "Epoch: 009/100 | Batch 500/600 | Cost: 0.2551\n",
      "Epoch: 009/100 | Batch 550/600 | Cost: 0.2049\n",
      "Epoch: 009/100 Train Cost: 0.2346\n",
      "Time elapsed: 2.06 min\n",
      "Epoch: 010/100 | Batch 000/600 | Cost: 0.1646\n",
      "Epoch: 010/100 | Batch 050/600 | Cost: 0.3388\n",
      "Epoch: 010/100 | Batch 100/600 | Cost: 0.1646\n",
      "Epoch: 010/100 | Batch 150/600 | Cost: 0.1914\n",
      "Epoch: 010/100 | Batch 200/600 | Cost: 0.3194\n",
      "Epoch: 010/100 | Batch 250/600 | Cost: 0.3007\n",
      "Epoch: 010/100 | Batch 300/600 | Cost: 0.2021\n",
      "Epoch: 010/100 | Batch 350/600 | Cost: 0.2315\n",
      "Epoch: 010/100 | Batch 400/600 | Cost: 0.1342\n",
      "Epoch: 010/100 | Batch 450/600 | Cost: 0.3294\n",
      "Epoch: 010/100 | Batch 500/600 | Cost: 0.1299\n",
      "Epoch: 010/100 | Batch 550/600 | Cost: 0.3791\n",
      "Epoch: 010/100 Train Cost: 0.2255\n",
      "Time elapsed: 2.29 min\n",
      "Epoch: 011/100 | Batch 000/600 | Cost: 0.1642\n",
      "Epoch: 011/100 | Batch 050/600 | Cost: 0.1817\n",
      "Epoch: 011/100 | Batch 100/600 | Cost: 0.2160\n",
      "Epoch: 011/100 | Batch 150/600 | Cost: 0.1853\n",
      "Epoch: 011/100 | Batch 200/600 | Cost: 0.1584\n",
      "Epoch: 011/100 | Batch 250/600 | Cost: 0.1958\n",
      "Epoch: 011/100 | Batch 300/600 | Cost: 0.2348\n",
      "Epoch: 011/100 | Batch 350/600 | Cost: 0.2308\n",
      "Epoch: 011/100 | Batch 400/600 | Cost: 0.2268\n",
      "Epoch: 011/100 | Batch 450/600 | Cost: 0.1644\n",
      "Epoch: 011/100 | Batch 500/600 | Cost: 0.2500\n",
      "Epoch: 011/100 | Batch 550/600 | Cost: 0.1751\n",
      "Epoch: 011/100 Train Cost: 0.2162\n",
      "Time elapsed: 2.51 min\n",
      "Epoch: 012/100 | Batch 000/600 | Cost: 0.1996\n",
      "Epoch: 012/100 | Batch 050/600 | Cost: 0.1668\n",
      "Epoch: 012/100 | Batch 100/600 | Cost: 0.1452\n",
      "Epoch: 012/100 | Batch 150/600 | Cost: 0.1781\n",
      "Epoch: 012/100 | Batch 200/600 | Cost: 0.2788\n",
      "Epoch: 012/100 | Batch 250/600 | Cost: 0.1862\n",
      "Epoch: 012/100 | Batch 300/600 | Cost: 0.0971\n",
      "Epoch: 012/100 | Batch 350/600 | Cost: 0.1960\n",
      "Epoch: 012/100 | Batch 400/600 | Cost: 0.3099\n",
      "Epoch: 012/100 | Batch 450/600 | Cost: 0.2870\n",
      "Epoch: 012/100 | Batch 500/600 | Cost: 0.2332\n",
      "Epoch: 012/100 | Batch 550/600 | Cost: 0.2117\n",
      "Epoch: 012/100 Train Cost: 0.2090\n",
      "Time elapsed: 2.74 min\n",
      "Epoch: 013/100 | Batch 000/600 | Cost: 0.1677\n",
      "Epoch: 013/100 | Batch 050/600 | Cost: 0.1206\n",
      "Epoch: 013/100 | Batch 100/600 | Cost: 0.1801\n",
      "Epoch: 013/100 | Batch 150/600 | Cost: 0.1496\n",
      "Epoch: 013/100 | Batch 200/600 | Cost: 0.2840\n",
      "Epoch: 013/100 | Batch 250/600 | Cost: 0.2171\n",
      "Epoch: 013/100 | Batch 300/600 | Cost: 0.1638\n",
      "Epoch: 013/100 | Batch 350/600 | Cost: 0.1197\n",
      "Epoch: 013/100 | Batch 400/600 | Cost: 0.2662\n",
      "Epoch: 013/100 | Batch 450/600 | Cost: 0.2606\n",
      "Epoch: 013/100 | Batch 500/600 | Cost: 0.1607\n",
      "Epoch: 013/100 | Batch 550/600 | Cost: 0.1412\n",
      "Epoch: 013/100 Train Cost: 0.2009\n",
      "Time elapsed: 2.97 min\n",
      "Epoch: 014/100 | Batch 000/600 | Cost: 0.1741\n",
      "Epoch: 014/100 | Batch 050/600 | Cost: 0.2499\n",
      "Epoch: 014/100 | Batch 100/600 | Cost: 0.1778\n",
      "Epoch: 014/100 | Batch 150/600 | Cost: 0.3010\n",
      "Epoch: 014/100 | Batch 200/600 | Cost: 0.2210\n",
      "Epoch: 014/100 | Batch 250/600 | Cost: 0.2179\n",
      "Epoch: 014/100 | Batch 300/600 | Cost: 0.1442\n",
      "Epoch: 014/100 | Batch 350/600 | Cost: 0.2517\n",
      "Epoch: 014/100 | Batch 400/600 | Cost: 0.2742\n",
      "Epoch: 014/100 | Batch 450/600 | Cost: 0.2662\n",
      "Epoch: 014/100 | Batch 500/600 | Cost: 0.2008\n",
      "Epoch: 014/100 | Batch 550/600 | Cost: 0.2900\n",
      "Epoch: 014/100 Train Cost: 0.1942\n",
      "Time elapsed: 3.20 min\n",
      "Epoch: 015/100 | Batch 000/600 | Cost: 0.1409\n",
      "Epoch: 015/100 | Batch 050/600 | Cost: 0.1476\n",
      "Epoch: 015/100 | Batch 100/600 | Cost: 0.2569\n",
      "Epoch: 015/100 | Batch 150/600 | Cost: 0.2905\n",
      "Epoch: 015/100 | Batch 200/600 | Cost: 0.1359\n",
      "Epoch: 015/100 | Batch 250/600 | Cost: 0.2022\n",
      "Epoch: 015/100 | Batch 300/600 | Cost: 0.1877\n",
      "Epoch: 015/100 | Batch 350/600 | Cost: 0.1849\n",
      "Epoch: 015/100 | Batch 400/600 | Cost: 0.1523\n",
      "Epoch: 015/100 | Batch 450/600 | Cost: 0.1554\n",
      "Epoch: 015/100 | Batch 500/600 | Cost: 0.2313\n",
      "Epoch: 015/100 | Batch 550/600 | Cost: 0.2300\n",
      "Epoch: 015/100 Train Cost: 0.1872\n",
      "Time elapsed: 3.44 min\n",
      "Epoch: 016/100 | Batch 000/600 | Cost: 0.1576\n",
      "Epoch: 016/100 | Batch 050/600 | Cost: 0.1234\n",
      "Epoch: 016/100 | Batch 100/600 | Cost: 0.1625\n",
      "Epoch: 016/100 | Batch 150/600 | Cost: 0.2896\n",
      "Epoch: 016/100 | Batch 200/600 | Cost: 0.1974\n",
      "Epoch: 016/100 | Batch 250/600 | Cost: 0.1342\n",
      "Epoch: 016/100 | Batch 300/600 | Cost: 0.1749\n",
      "Epoch: 016/100 | Batch 350/600 | Cost: 0.1141\n",
      "Epoch: 016/100 | Batch 400/600 | Cost: 0.1017\n",
      "Epoch: 016/100 | Batch 450/600 | Cost: 0.1160\n",
      "Epoch: 016/100 | Batch 500/600 | Cost: 0.1504\n",
      "Epoch: 016/100 | Batch 550/600 | Cost: 0.1416\n",
      "Epoch: 016/100 Train Cost: 0.1837\n",
      "Time elapsed: 3.67 min\n",
      "Epoch: 017/100 | Batch 000/600 | Cost: 0.1689\n",
      "Epoch: 017/100 | Batch 050/600 | Cost: 0.2447\n",
      "Epoch: 017/100 | Batch 100/600 | Cost: 0.2979\n",
      "Epoch: 017/100 | Batch 150/600 | Cost: 0.0968\n",
      "Epoch: 017/100 | Batch 200/600 | Cost: 0.0800\n",
      "Epoch: 017/100 | Batch 250/600 | Cost: 0.2950\n",
      "Epoch: 017/100 | Batch 300/600 | Cost: 0.2184\n",
      "Epoch: 017/100 | Batch 350/600 | Cost: 0.2086\n",
      "Epoch: 017/100 | Batch 400/600 | Cost: 0.2107\n",
      "Epoch: 017/100 | Batch 450/600 | Cost: 0.1347\n",
      "Epoch: 017/100 | Batch 500/600 | Cost: 0.1560\n",
      "Epoch: 017/100 | Batch 550/600 | Cost: 0.2639\n",
      "Epoch: 017/100 Train Cost: 0.1764\n",
      "Time elapsed: 3.90 min\n",
      "Epoch: 018/100 | Batch 000/600 | Cost: 0.1637\n",
      "Epoch: 018/100 | Batch 050/600 | Cost: 0.1792\n",
      "Epoch: 018/100 | Batch 100/600 | Cost: 0.1839\n",
      "Epoch: 018/100 | Batch 150/600 | Cost: 0.2327\n",
      "Epoch: 018/100 | Batch 200/600 | Cost: 0.2386\n",
      "Epoch: 018/100 | Batch 250/600 | Cost: 0.0920\n",
      "Epoch: 018/100 | Batch 300/600 | Cost: 0.2328\n",
      "Epoch: 018/100 | Batch 350/600 | Cost: 0.0785\n",
      "Epoch: 018/100 | Batch 400/600 | Cost: 0.1782\n",
      "Epoch: 018/100 | Batch 450/600 | Cost: 0.0779\n",
      "Epoch: 018/100 | Batch 500/600 | Cost: 0.1506\n",
      "Epoch: 018/100 | Batch 550/600 | Cost: 0.1625\n",
      "Epoch: 018/100 Train Cost: 0.1700\n",
      "Time elapsed: 4.13 min\n",
      "Epoch: 019/100 | Batch 000/600 | Cost: 0.2115\n",
      "Epoch: 019/100 | Batch 050/600 | Cost: 0.1803\n",
      "Epoch: 019/100 | Batch 100/600 | Cost: 0.2128\n",
      "Epoch: 019/100 | Batch 150/600 | Cost: 0.2429\n",
      "Epoch: 019/100 | Batch 200/600 | Cost: 0.1688\n",
      "Epoch: 019/100 | Batch 250/600 | Cost: 0.1088\n",
      "Epoch: 019/100 | Batch 300/600 | Cost: 0.0848\n",
      "Epoch: 019/100 | Batch 350/600 | Cost: 0.1217\n",
      "Epoch: 019/100 | Batch 400/600 | Cost: 0.1672\n",
      "Epoch: 019/100 | Batch 450/600 | Cost: 0.1344\n",
      "Epoch: 019/100 | Batch 500/600 | Cost: 0.2049\n",
      "Epoch: 019/100 | Batch 550/600 | Cost: 0.1028\n",
      "Epoch: 019/100 Train Cost: 0.1649\n",
      "Time elapsed: 4.36 min\n",
      "Epoch: 020/100 | Batch 000/600 | Cost: 0.1726\n",
      "Epoch: 020/100 | Batch 050/600 | Cost: 0.1760\n",
      "Epoch: 020/100 | Batch 100/600 | Cost: 0.1510\n",
      "Epoch: 020/100 | Batch 150/600 | Cost: 0.1237\n",
      "Epoch: 020/100 | Batch 200/600 | Cost: 0.1474\n",
      "Epoch: 020/100 | Batch 250/600 | Cost: 0.1175\n",
      "Epoch: 020/100 | Batch 300/600 | Cost: 0.2707\n",
      "Epoch: 020/100 | Batch 350/600 | Cost: 0.1088\n",
      "Epoch: 020/100 | Batch 400/600 | Cost: 0.1974\n",
      "Epoch: 020/100 | Batch 450/600 | Cost: 0.0967\n",
      "Epoch: 020/100 | Batch 500/600 | Cost: 0.1689\n",
      "Epoch: 020/100 | Batch 550/600 | Cost: 0.2246\n",
      "Epoch: 020/100 Train Cost: 0.1606\n",
      "Time elapsed: 4.56 min\n",
      "Epoch: 021/100 | Batch 000/600 | Cost: 0.1325\n",
      "Epoch: 021/100 | Batch 050/600 | Cost: 0.2344\n",
      "Epoch: 021/100 | Batch 100/600 | Cost: 0.1364\n",
      "Epoch: 021/100 | Batch 150/600 | Cost: 0.1347\n",
      "Epoch: 021/100 | Batch 200/600 | Cost: 0.1692\n",
      "Epoch: 021/100 | Batch 250/600 | Cost: 0.1744\n",
      "Epoch: 021/100 | Batch 300/600 | Cost: 0.1444\n",
      "Epoch: 021/100 | Batch 350/600 | Cost: 0.0504\n",
      "Epoch: 021/100 | Batch 400/600 | Cost: 0.1394\n",
      "Epoch: 021/100 | Batch 450/600 | Cost: 0.1177\n",
      "Epoch: 021/100 | Batch 500/600 | Cost: 0.1701\n",
      "Epoch: 021/100 | Batch 550/600 | Cost: 0.0691\n",
      "Epoch: 021/100 Train Cost: 0.1564\n",
      "Time elapsed: 4.77 min\n",
      "Epoch: 022/100 | Batch 000/600 | Cost: 0.1134\n",
      "Epoch: 022/100 | Batch 050/600 | Cost: 0.1648\n",
      "Epoch: 022/100 | Batch 100/600 | Cost: 0.1044\n",
      "Epoch: 022/100 | Batch 150/600 | Cost: 0.2775\n",
      "Epoch: 022/100 | Batch 200/600 | Cost: 0.1944\n",
      "Epoch: 022/100 | Batch 250/600 | Cost: 0.0902\n",
      "Epoch: 022/100 | Batch 300/600 | Cost: 0.1861\n",
      "Epoch: 022/100 | Batch 350/600 | Cost: 0.1615\n",
      "Epoch: 022/100 | Batch 400/600 | Cost: 0.0788\n",
      "Epoch: 022/100 | Batch 450/600 | Cost: 0.1062\n",
      "Epoch: 022/100 | Batch 500/600 | Cost: 0.1451\n",
      "Epoch: 022/100 | Batch 550/600 | Cost: 0.1139\n",
      "Epoch: 022/100 Train Cost: 0.1520\n",
      "Time elapsed: 4.97 min\n",
      "Epoch: 023/100 | Batch 000/600 | Cost: 0.0923\n",
      "Epoch: 023/100 | Batch 050/600 | Cost: 0.1927\n",
      "Epoch: 023/100 | Batch 100/600 | Cost: 0.1354\n",
      "Epoch: 023/100 | Batch 150/600 | Cost: 0.2150\n",
      "Epoch: 023/100 | Batch 200/600 | Cost: 0.1671\n",
      "Epoch: 023/100 | Batch 250/600 | Cost: 0.0854\n",
      "Epoch: 023/100 | Batch 300/600 | Cost: 0.1046\n",
      "Epoch: 023/100 | Batch 350/600 | Cost: 0.1435\n",
      "Epoch: 023/100 | Batch 400/600 | Cost: 0.2990\n",
      "Epoch: 023/100 | Batch 450/600 | Cost: 0.1602\n",
      "Epoch: 023/100 | Batch 500/600 | Cost: 0.0994\n",
      "Epoch: 023/100 | Batch 550/600 | Cost: 0.1363\n",
      "Epoch: 023/100 Train Cost: 0.1483\n",
      "Time elapsed: 5.17 min\n",
      "Epoch: 024/100 | Batch 000/600 | Cost: 0.1446\n",
      "Epoch: 024/100 | Batch 050/600 | Cost: 0.1877\n",
      "Epoch: 024/100 | Batch 100/600 | Cost: 0.1433\n",
      "Epoch: 024/100 | Batch 150/600 | Cost: 0.1949\n",
      "Epoch: 024/100 | Batch 200/600 | Cost: 0.1379\n",
      "Epoch: 024/100 | Batch 250/600 | Cost: 0.0825\n",
      "Epoch: 024/100 | Batch 300/600 | Cost: 0.1812\n",
      "Epoch: 024/100 | Batch 350/600 | Cost: 0.1702\n",
      "Epoch: 024/100 | Batch 400/600 | Cost: 0.1077\n",
      "Epoch: 024/100 | Batch 450/600 | Cost: 0.1393\n",
      "Epoch: 024/100 | Batch 500/600 | Cost: 0.0743\n",
      "Epoch: 024/100 | Batch 550/600 | Cost: 0.1864\n",
      "Epoch: 024/100 Train Cost: 0.1442\n",
      "Time elapsed: 5.37 min\n",
      "Epoch: 025/100 | Batch 000/600 | Cost: 0.1374\n",
      "Epoch: 025/100 | Batch 050/600 | Cost: 0.0816\n",
      "Epoch: 025/100 | Batch 100/600 | Cost: 0.1555\n",
      "Epoch: 025/100 | Batch 150/600 | Cost: 0.3188\n",
      "Epoch: 025/100 | Batch 200/600 | Cost: 0.0771\n",
      "Epoch: 025/100 | Batch 250/600 | Cost: 0.0710\n",
      "Epoch: 025/100 | Batch 300/600 | Cost: 0.1056\n",
      "Epoch: 025/100 | Batch 350/600 | Cost: 0.1342\n",
      "Epoch: 025/100 | Batch 400/600 | Cost: 0.1267\n",
      "Epoch: 025/100 | Batch 450/600 | Cost: 0.1132\n",
      "Epoch: 025/100 | Batch 500/600 | Cost: 0.0676\n",
      "Epoch: 025/100 | Batch 550/600 | Cost: 0.1746\n",
      "Epoch: 025/100 Train Cost: 0.1410\n",
      "Time elapsed: 5.58 min\n",
      "Epoch: 026/100 | Batch 000/600 | Cost: 0.2468\n",
      "Epoch: 026/100 | Batch 050/600 | Cost: 0.1016\n",
      "Epoch: 026/100 | Batch 100/600 | Cost: 0.0699\n",
      "Epoch: 026/100 | Batch 150/600 | Cost: 0.1265\n",
      "Epoch: 026/100 | Batch 200/600 | Cost: 0.0796\n",
      "Epoch: 026/100 | Batch 250/600 | Cost: 0.1604\n",
      "Epoch: 026/100 | Batch 300/600 | Cost: 0.1136\n",
      "Epoch: 026/100 | Batch 350/600 | Cost: 0.1026\n",
      "Epoch: 026/100 | Batch 400/600 | Cost: 0.2857\n",
      "Epoch: 026/100 | Batch 450/600 | Cost: 0.1369\n",
      "Epoch: 026/100 | Batch 500/600 | Cost: 0.1127\n",
      "Epoch: 026/100 | Batch 550/600 | Cost: 0.0652\n",
      "Epoch: 026/100 Train Cost: 0.1365\n",
      "Time elapsed: 5.78 min\n",
      "Epoch: 027/100 | Batch 000/600 | Cost: 0.1412\n",
      "Epoch: 027/100 | Batch 050/600 | Cost: 0.0775\n",
      "Epoch: 027/100 | Batch 100/600 | Cost: 0.1160\n",
      "Epoch: 027/100 | Batch 150/600 | Cost: 0.1008\n",
      "Epoch: 027/100 | Batch 200/600 | Cost: 0.1222\n",
      "Epoch: 027/100 | Batch 250/600 | Cost: 0.0875\n",
      "Epoch: 027/100 | Batch 300/600 | Cost: 0.0946\n",
      "Epoch: 027/100 | Batch 350/600 | Cost: 0.1971\n",
      "Epoch: 027/100 | Batch 400/600 | Cost: 0.2640\n",
      "Epoch: 027/100 | Batch 450/600 | Cost: 0.1502\n",
      "Epoch: 027/100 | Batch 500/600 | Cost: 0.1855\n",
      "Epoch: 027/100 | Batch 550/600 | Cost: 0.1752\n",
      "Epoch: 027/100 Train Cost: 0.1332\n",
      "Time elapsed: 5.98 min\n",
      "Epoch: 028/100 | Batch 000/600 | Cost: 0.1299\n",
      "Epoch: 028/100 | Batch 050/600 | Cost: 0.0864\n",
      "Epoch: 028/100 | Batch 100/600 | Cost: 0.1052\n",
      "Epoch: 028/100 | Batch 150/600 | Cost: 0.1704\n",
      "Epoch: 028/100 | Batch 200/600 | Cost: 0.1776\n",
      "Epoch: 028/100 | Batch 250/600 | Cost: 0.1226\n",
      "Epoch: 028/100 | Batch 300/600 | Cost: 0.1333\n",
      "Epoch: 028/100 | Batch 350/600 | Cost: 0.1991\n",
      "Epoch: 028/100 | Batch 400/600 | Cost: 0.1141\n",
      "Epoch: 028/100 | Batch 450/600 | Cost: 0.1308\n",
      "Epoch: 028/100 | Batch 500/600 | Cost: 0.1205\n",
      "Epoch: 028/100 | Batch 550/600 | Cost: 0.0789\n",
      "Epoch: 028/100 Train Cost: 0.1300\n",
      "Time elapsed: 6.19 min\n",
      "Epoch: 029/100 | Batch 000/600 | Cost: 0.0573\n",
      "Epoch: 029/100 | Batch 050/600 | Cost: 0.1613\n",
      "Epoch: 029/100 | Batch 100/600 | Cost: 0.1452\n",
      "Epoch: 029/100 | Batch 150/600 | Cost: 0.1496\n",
      "Epoch: 029/100 | Batch 200/600 | Cost: 0.1639\n",
      "Epoch: 029/100 | Batch 250/600 | Cost: 0.1110\n",
      "Epoch: 029/100 | Batch 300/600 | Cost: 0.0879\n",
      "Epoch: 029/100 | Batch 350/600 | Cost: 0.0762\n",
      "Epoch: 029/100 | Batch 400/600 | Cost: 0.1552\n",
      "Epoch: 029/100 | Batch 450/600 | Cost: 0.1567\n",
      "Epoch: 029/100 | Batch 500/600 | Cost: 0.1524\n",
      "Epoch: 029/100 | Batch 550/600 | Cost: 0.0880\n",
      "Epoch: 029/100 Train Cost: 0.1273\n",
      "Time elapsed: 6.40 min\n",
      "Epoch: 030/100 | Batch 000/600 | Cost: 0.0560\n",
      "Epoch: 030/100 | Batch 050/600 | Cost: 0.2088\n",
      "Epoch: 030/100 | Batch 100/600 | Cost: 0.0882\n",
      "Epoch: 030/100 | Batch 150/600 | Cost: 0.1265\n",
      "Epoch: 030/100 | Batch 200/600 | Cost: 0.1551\n",
      "Epoch: 030/100 | Batch 250/600 | Cost: 0.1086\n",
      "Epoch: 030/100 | Batch 300/600 | Cost: 0.1867\n",
      "Epoch: 030/100 | Batch 350/600 | Cost: 0.1490\n",
      "Epoch: 030/100 | Batch 400/600 | Cost: 0.1127\n",
      "Epoch: 030/100 | Batch 450/600 | Cost: 0.1354\n",
      "Epoch: 030/100 | Batch 500/600 | Cost: 0.2398\n",
      "Epoch: 030/100 | Batch 550/600 | Cost: 0.1404\n",
      "Epoch: 030/100 Train Cost: 0.1245\n",
      "Time elapsed: 6.61 min\n",
      "Epoch: 031/100 | Batch 000/600 | Cost: 0.1019\n",
      "Epoch: 031/100 | Batch 050/600 | Cost: 0.1525\n",
      "Epoch: 031/100 | Batch 100/600 | Cost: 0.2621\n",
      "Epoch: 031/100 | Batch 150/600 | Cost: 0.0958\n",
      "Epoch: 031/100 | Batch 200/600 | Cost: 0.0471\n",
      "Epoch: 031/100 | Batch 250/600 | Cost: 0.2023\n",
      "Epoch: 031/100 | Batch 300/600 | Cost: 0.0825\n",
      "Epoch: 031/100 | Batch 350/600 | Cost: 0.0635\n",
      "Epoch: 031/100 | Batch 400/600 | Cost: 0.1461\n",
      "Epoch: 031/100 | Batch 450/600 | Cost: 0.1381\n",
      "Epoch: 031/100 | Batch 500/600 | Cost: 0.1234\n",
      "Epoch: 031/100 | Batch 550/600 | Cost: 0.0957\n",
      "Epoch: 031/100 Train Cost: 0.1212\n",
      "Time elapsed: 6.81 min\n",
      "Epoch: 032/100 | Batch 000/600 | Cost: 0.2145\n",
      "Epoch: 032/100 | Batch 050/600 | Cost: 0.0856\n",
      "Epoch: 032/100 | Batch 100/600 | Cost: 0.1325\n",
      "Epoch: 032/100 | Batch 150/600 | Cost: 0.1943\n",
      "Epoch: 032/100 | Batch 200/600 | Cost: 0.1415\n",
      "Epoch: 032/100 | Batch 250/600 | Cost: 0.2055\n",
      "Epoch: 032/100 | Batch 300/600 | Cost: 0.1032\n",
      "Epoch: 032/100 | Batch 350/600 | Cost: 0.0801\n",
      "Epoch: 032/100 | Batch 400/600 | Cost: 0.0804\n",
      "Epoch: 032/100 | Batch 450/600 | Cost: 0.0944\n",
      "Epoch: 032/100 | Batch 500/600 | Cost: 0.0693\n",
      "Epoch: 032/100 | Batch 550/600 | Cost: 0.0480\n",
      "Epoch: 032/100 Train Cost: 0.1195\n",
      "Time elapsed: 7.02 min\n",
      "Epoch: 033/100 | Batch 000/600 | Cost: 0.1311\n",
      "Epoch: 033/100 | Batch 050/600 | Cost: 0.1033\n",
      "Epoch: 033/100 | Batch 100/600 | Cost: 0.1374\n",
      "Epoch: 033/100 | Batch 150/600 | Cost: 0.0616\n",
      "Epoch: 033/100 | Batch 200/600 | Cost: 0.1306\n",
      "Epoch: 033/100 | Batch 250/600 | Cost: 0.1751\n",
      "Epoch: 033/100 | Batch 300/600 | Cost: 0.2462\n",
      "Epoch: 033/100 | Batch 350/600 | Cost: 0.0716\n",
      "Epoch: 033/100 | Batch 400/600 | Cost: 0.0885\n",
      "Epoch: 033/100 | Batch 450/600 | Cost: 0.1359\n",
      "Epoch: 033/100 | Batch 500/600 | Cost: 0.1314\n",
      "Epoch: 033/100 | Batch 550/600 | Cost: 0.1589\n",
      "Epoch: 033/100 Train Cost: 0.1166\n",
      "Time elapsed: 7.22 min\n",
      "Epoch: 034/100 | Batch 000/600 | Cost: 0.1401\n",
      "Epoch: 034/100 | Batch 050/600 | Cost: 0.1449\n",
      "Epoch: 034/100 | Batch 100/600 | Cost: 0.1958\n",
      "Epoch: 034/100 | Batch 150/600 | Cost: 0.1579\n",
      "Epoch: 034/100 | Batch 200/600 | Cost: 0.1336\n",
      "Epoch: 034/100 | Batch 250/600 | Cost: 0.1109\n",
      "Epoch: 034/100 | Batch 300/600 | Cost: 0.1620\n",
      "Epoch: 034/100 | Batch 350/600 | Cost: 0.0917\n",
      "Epoch: 034/100 | Batch 400/600 | Cost: 0.1299\n",
      "Epoch: 034/100 | Batch 450/600 | Cost: 0.1775\n",
      "Epoch: 034/100 | Batch 500/600 | Cost: 0.1667\n",
      "Epoch: 034/100 | Batch 550/600 | Cost: 0.1281\n",
      "Epoch: 034/100 Train Cost: 0.1134\n",
      "Time elapsed: 7.42 min\n",
      "Epoch: 035/100 | Batch 000/600 | Cost: 0.1339\n",
      "Epoch: 035/100 | Batch 050/600 | Cost: 0.1088\n",
      "Epoch: 035/100 | Batch 100/600 | Cost: 0.1096\n",
      "Epoch: 035/100 | Batch 150/600 | Cost: 0.0954\n",
      "Epoch: 035/100 | Batch 200/600 | Cost: 0.0521\n",
      "Epoch: 035/100 | Batch 250/600 | Cost: 0.1042\n",
      "Epoch: 035/100 | Batch 300/600 | Cost: 0.1065\n",
      "Epoch: 035/100 | Batch 350/600 | Cost: 0.0700\n",
      "Epoch: 035/100 | Batch 400/600 | Cost: 0.1297\n",
      "Epoch: 035/100 | Batch 450/600 | Cost: 0.0803\n",
      "Epoch: 035/100 | Batch 500/600 | Cost: 0.0882\n",
      "Epoch: 035/100 | Batch 550/600 | Cost: 0.1897\n",
      "Epoch: 035/100 Train Cost: 0.1111\n",
      "Time elapsed: 7.62 min\n",
      "Epoch: 036/100 | Batch 000/600 | Cost: 0.0860\n",
      "Epoch: 036/100 | Batch 050/600 | Cost: 0.0903\n",
      "Epoch: 036/100 | Batch 100/600 | Cost: 0.0985\n",
      "Epoch: 036/100 | Batch 150/600 | Cost: 0.0849\n",
      "Epoch: 036/100 | Batch 200/600 | Cost: 0.0897\n",
      "Epoch: 036/100 | Batch 250/600 | Cost: 0.1243\n",
      "Epoch: 036/100 | Batch 300/600 | Cost: 0.1318\n",
      "Epoch: 036/100 | Batch 350/600 | Cost: 0.0890\n",
      "Epoch: 036/100 | Batch 400/600 | Cost: 0.1225\n",
      "Epoch: 036/100 | Batch 450/600 | Cost: 0.0872\n",
      "Epoch: 036/100 | Batch 500/600 | Cost: 0.1242\n",
      "Epoch: 036/100 | Batch 550/600 | Cost: 0.0594\n",
      "Epoch: 036/100 Train Cost: 0.1087\n",
      "Time elapsed: 7.83 min\n",
      "Epoch: 037/100 | Batch 000/600 | Cost: 0.1051\n",
      "Epoch: 037/100 | Batch 050/600 | Cost: 0.0709\n",
      "Epoch: 037/100 | Batch 100/600 | Cost: 0.1137\n",
      "Epoch: 037/100 | Batch 150/600 | Cost: 0.1999\n",
      "Epoch: 037/100 | Batch 200/600 | Cost: 0.0587\n",
      "Epoch: 037/100 | Batch 250/600 | Cost: 0.0637\n",
      "Epoch: 037/100 | Batch 300/600 | Cost: 0.0956\n",
      "Epoch: 037/100 | Batch 350/600 | Cost: 0.1742\n",
      "Epoch: 037/100 | Batch 400/600 | Cost: 0.0961\n",
      "Epoch: 037/100 | Batch 450/600 | Cost: 0.1114\n",
      "Epoch: 037/100 | Batch 500/600 | Cost: 0.0329\n",
      "Epoch: 037/100 | Batch 550/600 | Cost: 0.1418\n",
      "Epoch: 037/100 Train Cost: 0.1064\n",
      "Time elapsed: 8.03 min\n",
      "Epoch: 038/100 | Batch 000/600 | Cost: 0.2327\n",
      "Epoch: 038/100 | Batch 050/600 | Cost: 0.2211\n",
      "Epoch: 038/100 | Batch 100/600 | Cost: 0.0975\n",
      "Epoch: 038/100 | Batch 150/600 | Cost: 0.1153\n",
      "Epoch: 038/100 | Batch 200/600 | Cost: 0.0901\n",
      "Epoch: 038/100 | Batch 250/600 | Cost: 0.0694\n",
      "Epoch: 038/100 | Batch 300/600 | Cost: 0.0571\n",
      "Epoch: 038/100 | Batch 350/600 | Cost: 0.1350\n",
      "Epoch: 038/100 | Batch 400/600 | Cost: 0.0963\n",
      "Epoch: 038/100 | Batch 450/600 | Cost: 0.1809\n",
      "Epoch: 038/100 | Batch 500/600 | Cost: 0.0976\n",
      "Epoch: 038/100 | Batch 550/600 | Cost: 0.0852\n",
      "Epoch: 038/100 Train Cost: 0.1049\n",
      "Time elapsed: 8.23 min\n",
      "Epoch: 039/100 | Batch 000/600 | Cost: 0.1449\n",
      "Epoch: 039/100 | Batch 050/600 | Cost: 0.0931\n",
      "Epoch: 039/100 | Batch 100/600 | Cost: 0.0838\n",
      "Epoch: 039/100 | Batch 150/600 | Cost: 0.0791\n",
      "Epoch: 039/100 | Batch 200/600 | Cost: 0.0663\n",
      "Epoch: 039/100 | Batch 250/600 | Cost: 0.1186\n",
      "Epoch: 039/100 | Batch 300/600 | Cost: 0.1441\n",
      "Epoch: 039/100 | Batch 350/600 | Cost: 0.0581\n",
      "Epoch: 039/100 | Batch 400/600 | Cost: 0.0902\n",
      "Epoch: 039/100 | Batch 450/600 | Cost: 0.1833\n",
      "Epoch: 039/100 | Batch 500/600 | Cost: 0.1634\n",
      "Epoch: 039/100 | Batch 550/600 | Cost: 0.2217\n",
      "Epoch: 039/100 Train Cost: 0.1026\n",
      "Time elapsed: 8.43 min\n",
      "Epoch: 040/100 | Batch 000/600 | Cost: 0.1315\n",
      "Epoch: 040/100 | Batch 050/600 | Cost: 0.1522\n",
      "Epoch: 040/100 | Batch 100/600 | Cost: 0.0899\n",
      "Epoch: 040/100 | Batch 150/600 | Cost: 0.1340\n",
      "Epoch: 040/100 | Batch 200/600 | Cost: 0.1443\n",
      "Epoch: 040/100 | Batch 250/600 | Cost: 0.0343\n",
      "Epoch: 040/100 | Batch 300/600 | Cost: 0.0630\n",
      "Epoch: 040/100 | Batch 350/600 | Cost: 0.1529\n",
      "Epoch: 040/100 | Batch 400/600 | Cost: 0.1577\n",
      "Epoch: 040/100 | Batch 450/600 | Cost: 0.1038\n",
      "Epoch: 040/100 | Batch 500/600 | Cost: 0.0912\n",
      "Epoch: 040/100 | Batch 550/600 | Cost: 0.0362\n",
      "Epoch: 040/100 Train Cost: 0.1006\n",
      "Time elapsed: 8.64 min\n",
      "Epoch: 041/100 | Batch 000/600 | Cost: 0.0923\n",
      "Epoch: 041/100 | Batch 050/600 | Cost: 0.0985\n",
      "Epoch: 041/100 | Batch 100/600 | Cost: 0.0676\n",
      "Epoch: 041/100 | Batch 150/600 | Cost: 0.1105\n",
      "Epoch: 041/100 | Batch 200/600 | Cost: 0.0946\n",
      "Epoch: 041/100 | Batch 250/600 | Cost: 0.0996\n",
      "Epoch: 041/100 | Batch 300/600 | Cost: 0.0501\n",
      "Epoch: 041/100 | Batch 350/600 | Cost: 0.1240\n",
      "Epoch: 041/100 | Batch 400/600 | Cost: 0.1577\n",
      "Epoch: 041/100 | Batch 450/600 | Cost: 0.0569\n",
      "Epoch: 041/100 | Batch 500/600 | Cost: 0.1964\n",
      "Epoch: 041/100 | Batch 550/600 | Cost: 0.1088\n",
      "Epoch: 041/100 Train Cost: 0.0988\n",
      "Time elapsed: 8.84 min\n",
      "Epoch: 042/100 | Batch 000/600 | Cost: 0.1615\n",
      "Epoch: 042/100 | Batch 050/600 | Cost: 0.0973\n",
      "Epoch: 042/100 | Batch 100/600 | Cost: 0.1966\n",
      "Epoch: 042/100 | Batch 150/600 | Cost: 0.1062\n",
      "Epoch: 042/100 | Batch 200/600 | Cost: 0.0636\n",
      "Epoch: 042/100 | Batch 250/600 | Cost: 0.0962\n",
      "Epoch: 042/100 | Batch 300/600 | Cost: 0.1175\n",
      "Epoch: 042/100 | Batch 350/600 | Cost: 0.0632\n",
      "Epoch: 042/100 | Batch 400/600 | Cost: 0.0189\n",
      "Epoch: 042/100 | Batch 450/600 | Cost: 0.0907\n",
      "Epoch: 042/100 | Batch 500/600 | Cost: 0.0836\n",
      "Epoch: 042/100 | Batch 550/600 | Cost: 0.0461\n",
      "Epoch: 042/100 Train Cost: 0.0966\n",
      "Time elapsed: 9.05 min\n",
      "Epoch: 043/100 | Batch 000/600 | Cost: 0.1030\n",
      "Epoch: 043/100 | Batch 050/600 | Cost: 0.0493\n",
      "Epoch: 043/100 | Batch 100/600 | Cost: 0.0887\n",
      "Epoch: 043/100 | Batch 150/600 | Cost: 0.1068\n",
      "Epoch: 043/100 | Batch 200/600 | Cost: 0.0434\n",
      "Epoch: 043/100 | Batch 250/600 | Cost: 0.1217\n",
      "Epoch: 043/100 | Batch 300/600 | Cost: 0.0713\n",
      "Epoch: 043/100 | Batch 350/600 | Cost: 0.0932\n",
      "Epoch: 043/100 | Batch 400/600 | Cost: 0.0966\n",
      "Epoch: 043/100 | Batch 450/600 | Cost: 0.1086\n",
      "Epoch: 043/100 | Batch 500/600 | Cost: 0.1418\n",
      "Epoch: 043/100 | Batch 550/600 | Cost: 0.0772\n",
      "Epoch: 043/100 Train Cost: 0.0954\n",
      "Time elapsed: 9.25 min\n",
      "Epoch: 044/100 | Batch 000/600 | Cost: 0.1532\n",
      "Epoch: 044/100 | Batch 050/600 | Cost: 0.0401\n",
      "Epoch: 044/100 | Batch 100/600 | Cost: 0.1210\n",
      "Epoch: 044/100 | Batch 150/600 | Cost: 0.0442\n",
      "Epoch: 044/100 | Batch 200/600 | Cost: 0.0886\n",
      "Epoch: 044/100 | Batch 250/600 | Cost: 0.1620\n",
      "Epoch: 044/100 | Batch 300/600 | Cost: 0.0954\n",
      "Epoch: 044/100 | Batch 350/600 | Cost: 0.0934\n",
      "Epoch: 044/100 | Batch 400/600 | Cost: 0.0806\n",
      "Epoch: 044/100 | Batch 450/600 | Cost: 0.0933\n",
      "Epoch: 044/100 | Batch 500/600 | Cost: 0.1361\n",
      "Epoch: 044/100 | Batch 550/600 | Cost: 0.0306\n",
      "Epoch: 044/100 Train Cost: 0.0933\n",
      "Time elapsed: 9.46 min\n",
      "Epoch: 045/100 | Batch 000/600 | Cost: 0.1016\n",
      "Epoch: 045/100 | Batch 050/600 | Cost: 0.0854\n",
      "Epoch: 045/100 | Batch 100/600 | Cost: 0.1127\n",
      "Epoch: 045/100 | Batch 150/600 | Cost: 0.0744\n",
      "Epoch: 045/100 | Batch 200/600 | Cost: 0.0881\n",
      "Epoch: 045/100 | Batch 250/600 | Cost: 0.1330\n",
      "Epoch: 045/100 | Batch 300/600 | Cost: 0.0803\n",
      "Epoch: 045/100 | Batch 350/600 | Cost: 0.1047\n",
      "Epoch: 045/100 | Batch 400/600 | Cost: 0.1223\n",
      "Epoch: 045/100 | Batch 450/600 | Cost: 0.0529\n",
      "Epoch: 045/100 | Batch 500/600 | Cost: 0.2567\n",
      "Epoch: 045/100 | Batch 550/600 | Cost: 0.1231\n",
      "Epoch: 045/100 Train Cost: 0.0917\n",
      "Time elapsed: 9.67 min\n",
      "Epoch: 046/100 | Batch 000/600 | Cost: 0.1400\n",
      "Epoch: 046/100 | Batch 050/600 | Cost: 0.0389\n",
      "Epoch: 046/100 | Batch 100/600 | Cost: 0.1039\n",
      "Epoch: 046/100 | Batch 150/600 | Cost: 0.0802\n",
      "Epoch: 046/100 | Batch 200/600 | Cost: 0.0987\n",
      "Epoch: 046/100 | Batch 250/600 | Cost: 0.0959\n",
      "Epoch: 046/100 | Batch 300/600 | Cost: 0.0476\n",
      "Epoch: 046/100 | Batch 350/600 | Cost: 0.0915\n",
      "Epoch: 046/100 | Batch 400/600 | Cost: 0.1330\n",
      "Epoch: 046/100 | Batch 450/600 | Cost: 0.1129\n",
      "Epoch: 046/100 | Batch 500/600 | Cost: 0.0974\n",
      "Epoch: 046/100 | Batch 550/600 | Cost: 0.0573\n",
      "Epoch: 046/100 Train Cost: 0.0899\n",
      "Time elapsed: 9.87 min\n",
      "Epoch: 047/100 | Batch 000/600 | Cost: 0.0732\n",
      "Epoch: 047/100 | Batch 050/600 | Cost: 0.0441\n",
      "Epoch: 047/100 | Batch 100/600 | Cost: 0.0666\n",
      "Epoch: 047/100 | Batch 150/600 | Cost: 0.0997\n",
      "Epoch: 047/100 | Batch 200/600 | Cost: 0.0577\n",
      "Epoch: 047/100 | Batch 250/600 | Cost: 0.0460\n",
      "Epoch: 047/100 | Batch 300/600 | Cost: 0.1210\n",
      "Epoch: 047/100 | Batch 350/600 | Cost: 0.0636\n",
      "Epoch: 047/100 | Batch 400/600 | Cost: 0.0546\n",
      "Epoch: 047/100 | Batch 450/600 | Cost: 0.0629\n",
      "Epoch: 047/100 | Batch 500/600 | Cost: 0.1036\n",
      "Epoch: 047/100 | Batch 550/600 | Cost: 0.1527\n",
      "Epoch: 047/100 Train Cost: 0.0884\n",
      "Time elapsed: 10.08 min\n",
      "Epoch: 048/100 | Batch 000/600 | Cost: 0.1162\n",
      "Epoch: 048/100 | Batch 050/600 | Cost: 0.0434\n",
      "Epoch: 048/100 | Batch 100/600 | Cost: 0.0902\n",
      "Epoch: 048/100 | Batch 150/600 | Cost: 0.1385\n",
      "Epoch: 048/100 | Batch 200/600 | Cost: 0.0921\n",
      "Epoch: 048/100 | Batch 250/600 | Cost: 0.0567\n",
      "Epoch: 048/100 | Batch 300/600 | Cost: 0.0196\n",
      "Epoch: 048/100 | Batch 350/600 | Cost: 0.1726\n",
      "Epoch: 048/100 | Batch 400/600 | Cost: 0.1141\n",
      "Epoch: 048/100 | Batch 450/600 | Cost: 0.0742\n",
      "Epoch: 048/100 | Batch 500/600 | Cost: 0.1084\n",
      "Epoch: 048/100 | Batch 550/600 | Cost: 0.0900\n",
      "Epoch: 048/100 Train Cost: 0.0868\n",
      "Time elapsed: 10.28 min\n",
      "Epoch: 049/100 | Batch 000/600 | Cost: 0.0787\n",
      "Epoch: 049/100 | Batch 050/600 | Cost: 0.1052\n",
      "Epoch: 049/100 | Batch 100/600 | Cost: 0.1097\n",
      "Epoch: 049/100 | Batch 150/600 | Cost: 0.0539\n",
      "Epoch: 049/100 | Batch 200/600 | Cost: 0.0532\n",
      "Epoch: 049/100 | Batch 250/600 | Cost: 0.0602\n",
      "Epoch: 049/100 | Batch 300/600 | Cost: 0.0750\n",
      "Epoch: 049/100 | Batch 350/600 | Cost: 0.0880\n",
      "Epoch: 049/100 | Batch 400/600 | Cost: 0.1187\n",
      "Epoch: 049/100 | Batch 450/600 | Cost: 0.0752\n",
      "Epoch: 049/100 | Batch 500/600 | Cost: 0.0428\n",
      "Epoch: 049/100 | Batch 550/600 | Cost: 0.0562\n",
      "Epoch: 049/100 Train Cost: 0.0856\n",
      "Time elapsed: 10.49 min\n",
      "Epoch: 050/100 | Batch 000/600 | Cost: 0.0727\n",
      "Epoch: 050/100 | Batch 050/600 | Cost: 0.0506\n",
      "Epoch: 050/100 | Batch 100/600 | Cost: 0.0632\n",
      "Epoch: 050/100 | Batch 150/600 | Cost: 0.0255\n",
      "Epoch: 050/100 | Batch 200/600 | Cost: 0.1264\n",
      "Epoch: 050/100 | Batch 250/600 | Cost: 0.0863\n",
      "Epoch: 050/100 | Batch 300/600 | Cost: 0.1778\n",
      "Epoch: 050/100 | Batch 350/600 | Cost: 0.1720\n",
      "Epoch: 050/100 | Batch 400/600 | Cost: 0.1590\n",
      "Epoch: 050/100 | Batch 450/600 | Cost: 0.0815\n",
      "Epoch: 050/100 | Batch 500/600 | Cost: 0.0794\n",
      "Epoch: 050/100 | Batch 550/600 | Cost: 0.1488\n",
      "Epoch: 050/100 Train Cost: 0.0842\n",
      "Time elapsed: 10.70 min\n",
      "Epoch: 051/100 | Batch 000/600 | Cost: 0.0913\n",
      "Epoch: 051/100 | Batch 050/600 | Cost: 0.1693\n",
      "Epoch: 051/100 | Batch 100/600 | Cost: 0.0770\n",
      "Epoch: 051/100 | Batch 150/600 | Cost: 0.0400\n",
      "Epoch: 051/100 | Batch 200/600 | Cost: 0.0391\n",
      "Epoch: 051/100 | Batch 250/600 | Cost: 0.0716\n",
      "Epoch: 051/100 | Batch 300/600 | Cost: 0.0594\n",
      "Epoch: 051/100 | Batch 350/600 | Cost: 0.1243\n",
      "Epoch: 051/100 | Batch 400/600 | Cost: 0.0623\n",
      "Epoch: 051/100 | Batch 450/600 | Cost: 0.1595\n",
      "Epoch: 051/100 | Batch 500/600 | Cost: 0.1244\n",
      "Epoch: 051/100 | Batch 550/600 | Cost: 0.0726\n",
      "Epoch: 051/100 Train Cost: 0.0825\n",
      "Time elapsed: 10.90 min\n",
      "Epoch: 052/100 | Batch 000/600 | Cost: 0.0568\n",
      "Epoch: 052/100 | Batch 050/600 | Cost: 0.0757\n",
      "Epoch: 052/100 | Batch 100/600 | Cost: 0.0720\n",
      "Epoch: 052/100 | Batch 150/600 | Cost: 0.0353\n",
      "Epoch: 052/100 | Batch 200/600 | Cost: 0.0646\n",
      "Epoch: 052/100 | Batch 250/600 | Cost: 0.1003\n",
      "Epoch: 052/100 | Batch 300/600 | Cost: 0.0982\n",
      "Epoch: 052/100 | Batch 350/600 | Cost: 0.0683\n",
      "Epoch: 052/100 | Batch 400/600 | Cost: 0.1159\n",
      "Epoch: 052/100 | Batch 450/600 | Cost: 0.1553\n",
      "Epoch: 052/100 | Batch 500/600 | Cost: 0.0560\n",
      "Epoch: 052/100 | Batch 550/600 | Cost: 0.0710\n",
      "Epoch: 052/100 Train Cost: 0.0813\n",
      "Time elapsed: 11.11 min\n",
      "Epoch: 053/100 | Batch 000/600 | Cost: 0.1913\n",
      "Epoch: 053/100 | Batch 050/600 | Cost: 0.0623\n",
      "Epoch: 053/100 | Batch 100/600 | Cost: 0.1134\n",
      "Epoch: 053/100 | Batch 150/600 | Cost: 0.0999\n",
      "Epoch: 053/100 | Batch 200/600 | Cost: 0.0685\n",
      "Epoch: 053/100 | Batch 250/600 | Cost: 0.0887\n",
      "Epoch: 053/100 | Batch 300/600 | Cost: 0.0444\n",
      "Epoch: 053/100 | Batch 350/600 | Cost: 0.1293\n",
      "Epoch: 053/100 | Batch 400/600 | Cost: 0.0426\n",
      "Epoch: 053/100 | Batch 450/600 | Cost: 0.0453\n",
      "Epoch: 053/100 | Batch 500/600 | Cost: 0.1054\n",
      "Epoch: 053/100 | Batch 550/600 | Cost: 0.0836\n",
      "Epoch: 053/100 Train Cost: 0.0802\n",
      "Time elapsed: 11.31 min\n",
      "Epoch: 054/100 | Batch 000/600 | Cost: 0.0464\n",
      "Epoch: 054/100 | Batch 050/600 | Cost: 0.0463\n",
      "Epoch: 054/100 | Batch 100/600 | Cost: 0.1186\n",
      "Epoch: 054/100 | Batch 150/600 | Cost: 0.0368\n",
      "Epoch: 054/100 | Batch 200/600 | Cost: 0.1167\n",
      "Epoch: 054/100 | Batch 250/600 | Cost: 0.0396\n",
      "Epoch: 054/100 | Batch 300/600 | Cost: 0.0543\n",
      "Epoch: 054/100 | Batch 350/600 | Cost: 0.0307\n",
      "Epoch: 054/100 | Batch 400/600 | Cost: 0.0833\n",
      "Epoch: 054/100 | Batch 450/600 | Cost: 0.0462\n",
      "Epoch: 054/100 | Batch 500/600 | Cost: 0.0420\n",
      "Epoch: 054/100 | Batch 550/600 | Cost: 0.1046\n",
      "Epoch: 054/100 Train Cost: 0.0788\n",
      "Time elapsed: 11.52 min\n",
      "Epoch: 055/100 | Batch 000/600 | Cost: 0.0818\n",
      "Epoch: 055/100 | Batch 050/600 | Cost: 0.0773\n",
      "Epoch: 055/100 | Batch 100/600 | Cost: 0.0538\n",
      "Epoch: 055/100 | Batch 150/600 | Cost: 0.1075\n",
      "Epoch: 055/100 | Batch 200/600 | Cost: 0.0950\n",
      "Epoch: 055/100 | Batch 250/600 | Cost: 0.1132\n",
      "Epoch: 055/100 | Batch 300/600 | Cost: 0.0504\n",
      "Epoch: 055/100 | Batch 350/600 | Cost: 0.0884\n",
      "Epoch: 055/100 | Batch 400/600 | Cost: 0.0701\n",
      "Epoch: 055/100 | Batch 450/600 | Cost: 0.0956\n",
      "Epoch: 055/100 | Batch 500/600 | Cost: 0.0218\n",
      "Epoch: 055/100 | Batch 550/600 | Cost: 0.0709\n",
      "Epoch: 055/100 Train Cost: 0.0775\n",
      "Time elapsed: 11.72 min\n",
      "Epoch: 056/100 | Batch 000/600 | Cost: 0.0705\n",
      "Epoch: 056/100 | Batch 050/600 | Cost: 0.0731\n",
      "Epoch: 056/100 | Batch 100/600 | Cost: 0.0411\n",
      "Epoch: 056/100 | Batch 150/600 | Cost: 0.0716\n",
      "Epoch: 056/100 | Batch 200/600 | Cost: 0.0313\n",
      "Epoch: 056/100 | Batch 250/600 | Cost: 0.1265\n",
      "Epoch: 056/100 | Batch 300/600 | Cost: 0.0826\n",
      "Epoch: 056/100 | Batch 350/600 | Cost: 0.0421\n",
      "Epoch: 056/100 | Batch 400/600 | Cost: 0.0459\n",
      "Epoch: 056/100 | Batch 450/600 | Cost: 0.0657\n",
      "Epoch: 056/100 | Batch 500/600 | Cost: 0.0851\n",
      "Epoch: 056/100 | Batch 550/600 | Cost: 0.0893\n",
      "Epoch: 056/100 Train Cost: 0.0762\n",
      "Time elapsed: 11.93 min\n",
      "Epoch: 057/100 | Batch 000/600 | Cost: 0.0575\n",
      "Epoch: 057/100 | Batch 050/600 | Cost: 0.0817\n",
      "Epoch: 057/100 | Batch 100/600 | Cost: 0.1459\n",
      "Epoch: 057/100 | Batch 150/600 | Cost: 0.0753\n",
      "Epoch: 057/100 | Batch 200/600 | Cost: 0.1062\n",
      "Epoch: 057/100 | Batch 250/600 | Cost: 0.0407\n",
      "Epoch: 057/100 | Batch 300/600 | Cost: 0.0633\n",
      "Epoch: 057/100 | Batch 350/600 | Cost: 0.0328\n",
      "Epoch: 057/100 | Batch 400/600 | Cost: 0.1291\n",
      "Epoch: 057/100 | Batch 450/600 | Cost: 0.0342\n",
      "Epoch: 057/100 | Batch 500/600 | Cost: 0.0218\n",
      "Epoch: 057/100 | Batch 550/600 | Cost: 0.0584\n",
      "Epoch: 057/100 Train Cost: 0.0751\n",
      "Time elapsed: 12.14 min\n",
      "Epoch: 058/100 | Batch 000/600 | Cost: 0.0417\n",
      "Epoch: 058/100 | Batch 050/600 | Cost: 0.0565\n",
      "Epoch: 058/100 | Batch 100/600 | Cost: 0.0399\n",
      "Epoch: 058/100 | Batch 150/600 | Cost: 0.1014\n",
      "Epoch: 058/100 | Batch 200/600 | Cost: 0.1079\n",
      "Epoch: 058/100 | Batch 250/600 | Cost: 0.1500\n",
      "Epoch: 058/100 | Batch 300/600 | Cost: 0.1033\n",
      "Epoch: 058/100 | Batch 350/600 | Cost: 0.0419\n",
      "Epoch: 058/100 | Batch 400/600 | Cost: 0.0514\n",
      "Epoch: 058/100 | Batch 450/600 | Cost: 0.1354\n",
      "Epoch: 058/100 | Batch 500/600 | Cost: 0.0949\n",
      "Epoch: 058/100 | Batch 550/600 | Cost: 0.0540\n",
      "Epoch: 058/100 Train Cost: 0.0738\n",
      "Time elapsed: 12.35 min\n",
      "Epoch: 059/100 | Batch 000/600 | Cost: 0.0506\n",
      "Epoch: 059/100 | Batch 050/600 | Cost: 0.0407\n",
      "Epoch: 059/100 | Batch 100/600 | Cost: 0.0747\n",
      "Epoch: 059/100 | Batch 150/600 | Cost: 0.0495\n",
      "Epoch: 059/100 | Batch 200/600 | Cost: 0.0387\n",
      "Epoch: 059/100 | Batch 250/600 | Cost: 0.0588\n",
      "Epoch: 059/100 | Batch 300/600 | Cost: 0.0171\n",
      "Epoch: 059/100 | Batch 350/600 | Cost: 0.0463\n",
      "Epoch: 059/100 | Batch 400/600 | Cost: 0.1268\n",
      "Epoch: 059/100 | Batch 450/600 | Cost: 0.0483\n",
      "Epoch: 059/100 | Batch 500/600 | Cost: 0.0385\n",
      "Epoch: 059/100 | Batch 550/600 | Cost: 0.0614\n",
      "Epoch: 059/100 Train Cost: 0.0728\n",
      "Time elapsed: 12.55 min\n",
      "Epoch: 060/100 | Batch 000/600 | Cost: 0.0701\n",
      "Epoch: 060/100 | Batch 050/600 | Cost: 0.0664\n",
      "Epoch: 060/100 | Batch 100/600 | Cost: 0.1066\n",
      "Epoch: 060/100 | Batch 150/600 | Cost: 0.0364\n",
      "Epoch: 060/100 | Batch 200/600 | Cost: 0.0826\n",
      "Epoch: 060/100 | Batch 250/600 | Cost: 0.1059\n",
      "Epoch: 060/100 | Batch 300/600 | Cost: 0.1053\n",
      "Epoch: 060/100 | Batch 350/600 | Cost: 0.0955\n",
      "Epoch: 060/100 | Batch 400/600 | Cost: 0.0543\n",
      "Epoch: 060/100 | Batch 450/600 | Cost: 0.0765\n",
      "Epoch: 060/100 | Batch 500/600 | Cost: 0.1035\n",
      "Epoch: 060/100 | Batch 550/600 | Cost: 0.0973\n",
      "Epoch: 060/100 Train Cost: 0.0718\n",
      "Time elapsed: 12.76 min\n",
      "Epoch: 061/100 | Batch 000/600 | Cost: 0.0809\n",
      "Epoch: 061/100 | Batch 050/600 | Cost: 0.1285\n",
      "Epoch: 061/100 | Batch 100/600 | Cost: 0.0842\n",
      "Epoch: 061/100 | Batch 150/600 | Cost: 0.1137\n",
      "Epoch: 061/100 | Batch 200/600 | Cost: 0.0620\n",
      "Epoch: 061/100 | Batch 250/600 | Cost: 0.0719\n",
      "Epoch: 061/100 | Batch 300/600 | Cost: 0.0649\n",
      "Epoch: 061/100 | Batch 350/600 | Cost: 0.0257\n",
      "Epoch: 061/100 | Batch 400/600 | Cost: 0.0266\n",
      "Epoch: 061/100 | Batch 450/600 | Cost: 0.0512\n",
      "Epoch: 061/100 | Batch 500/600 | Cost: 0.0760\n",
      "Epoch: 061/100 | Batch 550/600 | Cost: 0.0526\n",
      "Epoch: 061/100 Train Cost: 0.0708\n",
      "Time elapsed: 12.97 min\n",
      "Epoch: 062/100 | Batch 000/600 | Cost: 0.0296\n",
      "Epoch: 062/100 | Batch 050/600 | Cost: 0.0583\n",
      "Epoch: 062/100 | Batch 100/600 | Cost: 0.0520\n",
      "Epoch: 062/100 | Batch 150/600 | Cost: 0.0292\n",
      "Epoch: 062/100 | Batch 200/600 | Cost: 0.0722\n",
      "Epoch: 062/100 | Batch 250/600 | Cost: 0.0385\n",
      "Epoch: 062/100 | Batch 300/600 | Cost: 0.0510\n",
      "Epoch: 062/100 | Batch 350/600 | Cost: 0.1137\n",
      "Epoch: 062/100 | Batch 400/600 | Cost: 0.0503\n",
      "Epoch: 062/100 | Batch 450/600 | Cost: 0.0625\n",
      "Epoch: 062/100 | Batch 500/600 | Cost: 0.0359\n",
      "Epoch: 062/100 | Batch 550/600 | Cost: 0.1604\n",
      "Epoch: 062/100 Train Cost: 0.0701\n",
      "Time elapsed: 13.18 min\n",
      "Epoch: 063/100 | Batch 000/600 | Cost: 0.0776\n",
      "Epoch: 063/100 | Batch 050/600 | Cost: 0.0847\n",
      "Epoch: 063/100 | Batch 100/600 | Cost: 0.0855\n",
      "Epoch: 063/100 | Batch 150/600 | Cost: 0.0659\n",
      "Epoch: 063/100 | Batch 200/600 | Cost: 0.0380\n",
      "Epoch: 063/100 | Batch 250/600 | Cost: 0.0559\n",
      "Epoch: 063/100 | Batch 300/600 | Cost: 0.1015\n",
      "Epoch: 063/100 | Batch 350/600 | Cost: 0.1206\n",
      "Epoch: 063/100 | Batch 400/600 | Cost: 0.0567\n",
      "Epoch: 063/100 | Batch 450/600 | Cost: 0.0222\n",
      "Epoch: 063/100 | Batch 500/600 | Cost: 0.0957\n",
      "Epoch: 063/100 | Batch 550/600 | Cost: 0.0475\n",
      "Epoch: 063/100 Train Cost: 0.0688\n",
      "Time elapsed: 13.39 min\n",
      "Epoch: 064/100 | Batch 000/600 | Cost: 0.0433\n",
      "Epoch: 064/100 | Batch 050/600 | Cost: 0.1092\n",
      "Epoch: 064/100 | Batch 100/600 | Cost: 0.0714\n",
      "Epoch: 064/100 | Batch 150/600 | Cost: 0.1441\n",
      "Epoch: 064/100 | Batch 200/600 | Cost: 0.0688\n",
      "Epoch: 064/100 | Batch 250/600 | Cost: 0.0942\n",
      "Epoch: 064/100 | Batch 300/600 | Cost: 0.1130\n",
      "Epoch: 064/100 | Batch 350/600 | Cost: 0.0891\n",
      "Epoch: 064/100 | Batch 400/600 | Cost: 0.1555\n",
      "Epoch: 064/100 | Batch 450/600 | Cost: 0.0602\n",
      "Epoch: 064/100 | Batch 500/600 | Cost: 0.0335\n",
      "Epoch: 064/100 | Batch 550/600 | Cost: 0.0396\n",
      "Epoch: 064/100 Train Cost: 0.0677\n",
      "Time elapsed: 13.60 min\n",
      "Epoch: 065/100 | Batch 000/600 | Cost: 0.1196\n",
      "Epoch: 065/100 | Batch 050/600 | Cost: 0.0201\n",
      "Epoch: 065/100 | Batch 100/600 | Cost: 0.0887\n",
      "Epoch: 065/100 | Batch 150/600 | Cost: 0.0779\n",
      "Epoch: 065/100 | Batch 200/600 | Cost: 0.0920\n",
      "Epoch: 065/100 | Batch 250/600 | Cost: 0.0433\n",
      "Epoch: 065/100 | Batch 300/600 | Cost: 0.1385\n",
      "Epoch: 065/100 | Batch 350/600 | Cost: 0.0750\n",
      "Epoch: 065/100 | Batch 400/600 | Cost: 0.1172\n",
      "Epoch: 065/100 | Batch 450/600 | Cost: 0.0524\n",
      "Epoch: 065/100 | Batch 500/600 | Cost: 0.0590\n",
      "Epoch: 065/100 | Batch 550/600 | Cost: 0.1759\n",
      "Epoch: 065/100 Train Cost: 0.0666\n",
      "Time elapsed: 13.81 min\n",
      "Epoch: 066/100 | Batch 000/600 | Cost: 0.0391\n",
      "Epoch: 066/100 | Batch 050/600 | Cost: 0.0425\n",
      "Epoch: 066/100 | Batch 100/600 | Cost: 0.1042\n",
      "Epoch: 066/100 | Batch 150/600 | Cost: 0.1636\n",
      "Epoch: 066/100 | Batch 200/600 | Cost: 0.0739\n",
      "Epoch: 066/100 | Batch 250/600 | Cost: 0.0708\n",
      "Epoch: 066/100 | Batch 300/600 | Cost: 0.0799\n",
      "Epoch: 066/100 | Batch 350/600 | Cost: 0.0285\n",
      "Epoch: 066/100 | Batch 400/600 | Cost: 0.0368\n",
      "Epoch: 066/100 | Batch 450/600 | Cost: 0.0140\n",
      "Epoch: 066/100 | Batch 500/600 | Cost: 0.0446\n",
      "Epoch: 066/100 | Batch 550/600 | Cost: 0.1237\n",
      "Epoch: 066/100 Train Cost: 0.0657\n",
      "Time elapsed: 14.01 min\n",
      "Epoch: 067/100 | Batch 000/600 | Cost: 0.0528\n",
      "Epoch: 067/100 | Batch 050/600 | Cost: 0.0429\n",
      "Epoch: 067/100 | Batch 100/600 | Cost: 0.0703\n",
      "Epoch: 067/100 | Batch 150/600 | Cost: 0.0703\n",
      "Epoch: 067/100 | Batch 200/600 | Cost: 0.0882\n",
      "Epoch: 067/100 | Batch 250/600 | Cost: 0.1066\n",
      "Epoch: 067/100 | Batch 300/600 | Cost: 0.1619\n",
      "Epoch: 067/100 | Batch 350/600 | Cost: 0.0356\n",
      "Epoch: 067/100 | Batch 400/600 | Cost: 0.1049\n",
      "Epoch: 067/100 | Batch 450/600 | Cost: 0.0461\n",
      "Epoch: 067/100 | Batch 500/600 | Cost: 0.0434\n",
      "Epoch: 067/100 | Batch 550/600 | Cost: 0.0443\n",
      "Epoch: 067/100 Train Cost: 0.0650\n",
      "Time elapsed: 14.22 min\n",
      "Epoch: 068/100 | Batch 000/600 | Cost: 0.1043\n",
      "Epoch: 068/100 | Batch 050/600 | Cost: 0.0514\n",
      "Epoch: 068/100 | Batch 100/600 | Cost: 0.0404\n",
      "Epoch: 068/100 | Batch 150/600 | Cost: 0.0359\n",
      "Epoch: 068/100 | Batch 200/600 | Cost: 0.1113\n",
      "Epoch: 068/100 | Batch 250/600 | Cost: 0.0583\n",
      "Epoch: 068/100 | Batch 300/600 | Cost: 0.0530\n",
      "Epoch: 068/100 | Batch 350/600 | Cost: 0.0252\n",
      "Epoch: 068/100 | Batch 400/600 | Cost: 0.1109\n",
      "Epoch: 068/100 | Batch 450/600 | Cost: 0.0342\n",
      "Epoch: 068/100 | Batch 500/600 | Cost: 0.0748\n",
      "Epoch: 068/100 | Batch 550/600 | Cost: 0.0807\n",
      "Epoch: 068/100 Train Cost: 0.0640\n",
      "Time elapsed: 14.43 min\n",
      "Epoch: 069/100 | Batch 000/600 | Cost: 0.1013\n",
      "Epoch: 069/100 | Batch 050/600 | Cost: 0.0346\n",
      "Epoch: 069/100 | Batch 100/600 | Cost: 0.0658\n",
      "Epoch: 069/100 | Batch 150/600 | Cost: 0.0124\n",
      "Epoch: 069/100 | Batch 200/600 | Cost: 0.0662\n",
      "Epoch: 069/100 | Batch 250/600 | Cost: 0.0463\n",
      "Epoch: 069/100 | Batch 300/600 | Cost: 0.1318\n",
      "Epoch: 069/100 | Batch 350/600 | Cost: 0.1029\n",
      "Epoch: 069/100 | Batch 400/600 | Cost: 0.0644\n",
      "Epoch: 069/100 | Batch 450/600 | Cost: 0.0839\n",
      "Epoch: 069/100 | Batch 500/600 | Cost: 0.0568\n",
      "Epoch: 069/100 | Batch 550/600 | Cost: 0.0658\n",
      "Epoch: 069/100 Train Cost: 0.0634\n",
      "Time elapsed: 14.64 min\n",
      "Epoch: 070/100 | Batch 000/600 | Cost: 0.0249\n",
      "Epoch: 070/100 | Batch 050/600 | Cost: 0.0501\n",
      "Epoch: 070/100 | Batch 100/600 | Cost: 0.1417\n",
      "Epoch: 070/100 | Batch 150/600 | Cost: 0.0611\n",
      "Epoch: 070/100 | Batch 200/600 | Cost: 0.0525\n",
      "Epoch: 070/100 | Batch 250/600 | Cost: 0.0177\n",
      "Epoch: 070/100 | Batch 300/600 | Cost: 0.0494\n",
      "Epoch: 070/100 | Batch 350/600 | Cost: 0.0437\n",
      "Epoch: 070/100 | Batch 400/600 | Cost: 0.0559\n",
      "Epoch: 070/100 | Batch 450/600 | Cost: 0.0245\n",
      "Epoch: 070/100 | Batch 500/600 | Cost: 0.0559\n",
      "Epoch: 070/100 | Batch 550/600 | Cost: 0.0225\n",
      "Epoch: 070/100 Train Cost: 0.0621\n",
      "Time elapsed: 14.85 min\n",
      "Epoch: 071/100 | Batch 000/600 | Cost: 0.1073\n",
      "Epoch: 071/100 | Batch 050/600 | Cost: 0.0725\n",
      "Epoch: 071/100 | Batch 100/600 | Cost: 0.0202\n",
      "Epoch: 071/100 | Batch 150/600 | Cost: 0.0338\n",
      "Epoch: 071/100 | Batch 200/600 | Cost: 0.0380\n",
      "Epoch: 071/100 | Batch 250/600 | Cost: 0.1082\n",
      "Epoch: 071/100 | Batch 300/600 | Cost: 0.0727\n",
      "Epoch: 071/100 | Batch 350/600 | Cost: 0.0345\n",
      "Epoch: 071/100 | Batch 400/600 | Cost: 0.0385\n",
      "Epoch: 071/100 | Batch 450/600 | Cost: 0.0866\n",
      "Epoch: 071/100 | Batch 500/600 | Cost: 0.0525\n",
      "Epoch: 071/100 | Batch 550/600 | Cost: 0.1330\n",
      "Epoch: 071/100 Train Cost: 0.0615\n",
      "Time elapsed: 15.06 min\n",
      "Epoch: 072/100 | Batch 000/600 | Cost: 0.0995\n",
      "Epoch: 072/100 | Batch 050/600 | Cost: 0.0361\n",
      "Epoch: 072/100 | Batch 100/600 | Cost: 0.1166\n",
      "Epoch: 072/100 | Batch 150/600 | Cost: 0.0713\n",
      "Epoch: 072/100 | Batch 200/600 | Cost: 0.0755\n",
      "Epoch: 072/100 | Batch 250/600 | Cost: 0.0368\n",
      "Epoch: 072/100 | Batch 300/600 | Cost: 0.0415\n",
      "Epoch: 072/100 | Batch 350/600 | Cost: 0.0484\n",
      "Epoch: 072/100 | Batch 400/600 | Cost: 0.0415\n",
      "Epoch: 072/100 | Batch 450/600 | Cost: 0.0707\n",
      "Epoch: 072/100 | Batch 500/600 | Cost: 0.0618\n",
      "Epoch: 072/100 | Batch 550/600 | Cost: 0.0464\n",
      "Epoch: 072/100 Train Cost: 0.0608\n",
      "Time elapsed: 15.27 min\n",
      "Epoch: 073/100 | Batch 000/600 | Cost: 0.0393\n",
      "Epoch: 073/100 | Batch 050/600 | Cost: 0.0292\n",
      "Epoch: 073/100 | Batch 100/600 | Cost: 0.0263\n",
      "Epoch: 073/100 | Batch 150/600 | Cost: 0.0920\n",
      "Epoch: 073/100 | Batch 200/600 | Cost: 0.0725\n",
      "Epoch: 073/100 | Batch 250/600 | Cost: 0.0749\n",
      "Epoch: 073/100 | Batch 300/600 | Cost: 0.0424\n",
      "Epoch: 073/100 | Batch 350/600 | Cost: 0.0424\n",
      "Epoch: 073/100 | Batch 400/600 | Cost: 0.0232\n",
      "Epoch: 073/100 | Batch 450/600 | Cost: 0.0587\n",
      "Epoch: 073/100 | Batch 500/600 | Cost: 0.0552\n",
      "Epoch: 073/100 | Batch 550/600 | Cost: 0.0170\n",
      "Epoch: 073/100 Train Cost: 0.0598\n",
      "Time elapsed: 15.48 min\n",
      "Epoch: 074/100 | Batch 000/600 | Cost: 0.0560\n",
      "Epoch: 074/100 | Batch 050/600 | Cost: 0.0392\n",
      "Epoch: 074/100 | Batch 100/600 | Cost: 0.0510\n",
      "Epoch: 074/100 | Batch 150/600 | Cost: 0.0542\n",
      "Epoch: 074/100 | Batch 200/600 | Cost: 0.0543\n",
      "Epoch: 074/100 | Batch 250/600 | Cost: 0.0447\n",
      "Epoch: 074/100 | Batch 300/600 | Cost: 0.0376\n",
      "Epoch: 074/100 | Batch 350/600 | Cost: 0.0281\n",
      "Epoch: 074/100 | Batch 400/600 | Cost: 0.0982\n",
      "Epoch: 074/100 | Batch 450/600 | Cost: 0.0454\n",
      "Epoch: 074/100 | Batch 500/600 | Cost: 0.0646\n",
      "Epoch: 074/100 | Batch 550/600 | Cost: 0.0367\n",
      "Epoch: 074/100 Train Cost: 0.0590\n",
      "Time elapsed: 15.68 min\n",
      "Epoch: 075/100 | Batch 000/600 | Cost: 0.0802\n",
      "Epoch: 075/100 | Batch 050/600 | Cost: 0.0493\n",
      "Epoch: 075/100 | Batch 100/600 | Cost: 0.1200\n",
      "Epoch: 075/100 | Batch 150/600 | Cost: 0.0518\n",
      "Epoch: 075/100 | Batch 200/600 | Cost: 0.0332\n",
      "Epoch: 075/100 | Batch 250/600 | Cost: 0.0193\n",
      "Epoch: 075/100 | Batch 300/600 | Cost: 0.0746\n",
      "Epoch: 075/100 | Batch 350/600 | Cost: 0.1338\n",
      "Epoch: 075/100 | Batch 400/600 | Cost: 0.0210\n",
      "Epoch: 075/100 | Batch 450/600 | Cost: 0.0798\n",
      "Epoch: 075/100 | Batch 500/600 | Cost: 0.0461\n",
      "Epoch: 075/100 | Batch 550/600 | Cost: 0.0224\n",
      "Epoch: 075/100 Train Cost: 0.0584\n",
      "Time elapsed: 15.89 min\n",
      "Epoch: 076/100 | Batch 000/600 | Cost: 0.0736\n",
      "Epoch: 076/100 | Batch 050/600 | Cost: 0.0677\n",
      "Epoch: 076/100 | Batch 100/600 | Cost: 0.0974\n",
      "Epoch: 076/100 | Batch 150/600 | Cost: 0.0288\n",
      "Epoch: 076/100 | Batch 200/600 | Cost: 0.0968\n",
      "Epoch: 076/100 | Batch 250/600 | Cost: 0.0743\n",
      "Epoch: 076/100 | Batch 300/600 | Cost: 0.0384\n",
      "Epoch: 076/100 | Batch 350/600 | Cost: 0.0663\n",
      "Epoch: 076/100 | Batch 400/600 | Cost: 0.0319\n",
      "Epoch: 076/100 | Batch 450/600 | Cost: 0.0286\n",
      "Epoch: 076/100 | Batch 500/600 | Cost: 0.0599\n",
      "Epoch: 076/100 | Batch 550/600 | Cost: 0.0321\n",
      "Epoch: 076/100 Train Cost: 0.0575\n",
      "Time elapsed: 16.09 min\n",
      "Epoch: 077/100 | Batch 000/600 | Cost: 0.0332\n",
      "Epoch: 077/100 | Batch 050/600 | Cost: 0.0749\n",
      "Epoch: 077/100 | Batch 100/600 | Cost: 0.0338\n",
      "Epoch: 077/100 | Batch 150/600 | Cost: 0.0287\n",
      "Epoch: 077/100 | Batch 200/600 | Cost: 0.0508\n",
      "Epoch: 077/100 | Batch 250/600 | Cost: 0.1056\n",
      "Epoch: 077/100 | Batch 300/600 | Cost: 0.0855\n",
      "Epoch: 077/100 | Batch 350/600 | Cost: 0.0388\n",
      "Epoch: 077/100 | Batch 400/600 | Cost: 0.0882\n",
      "Epoch: 077/100 | Batch 450/600 | Cost: 0.0259\n",
      "Epoch: 077/100 | Batch 500/600 | Cost: 0.0509\n",
      "Epoch: 077/100 | Batch 550/600 | Cost: 0.0511\n",
      "Epoch: 077/100 Train Cost: 0.0568\n",
      "Time elapsed: 16.30 min\n",
      "Epoch: 078/100 | Batch 000/600 | Cost: 0.0204\n",
      "Epoch: 078/100 | Batch 050/600 | Cost: 0.0352\n",
      "Epoch: 078/100 | Batch 100/600 | Cost: 0.0307\n",
      "Epoch: 078/100 | Batch 150/600 | Cost: 0.0294\n",
      "Epoch: 078/100 | Batch 200/600 | Cost: 0.0447\n",
      "Epoch: 078/100 | Batch 250/600 | Cost: 0.0082\n",
      "Epoch: 078/100 | Batch 300/600 | Cost: 0.0402\n",
      "Epoch: 078/100 | Batch 350/600 | Cost: 0.0825\n",
      "Epoch: 078/100 | Batch 400/600 | Cost: 0.0609\n",
      "Epoch: 078/100 | Batch 450/600 | Cost: 0.0284\n",
      "Epoch: 078/100 | Batch 500/600 | Cost: 0.0212\n",
      "Epoch: 078/100 | Batch 550/600 | Cost: 0.1203\n",
      "Epoch: 078/100 Train Cost: 0.0560\n",
      "Time elapsed: 16.51 min\n",
      "Epoch: 079/100 | Batch 000/600 | Cost: 0.0687\n",
      "Epoch: 079/100 | Batch 050/600 | Cost: 0.0587\n",
      "Epoch: 079/100 | Batch 100/600 | Cost: 0.0314\n",
      "Epoch: 079/100 | Batch 150/600 | Cost: 0.0765\n",
      "Epoch: 079/100 | Batch 200/600 | Cost: 0.0680\n",
      "Epoch: 079/100 | Batch 250/600 | Cost: 0.0872\n",
      "Epoch: 079/100 | Batch 300/600 | Cost: 0.0359\n",
      "Epoch: 079/100 | Batch 350/600 | Cost: 0.0936\n",
      "Epoch: 079/100 | Batch 400/600 | Cost: 0.1070\n",
      "Epoch: 079/100 | Batch 450/600 | Cost: 0.0293\n",
      "Epoch: 079/100 | Batch 500/600 | Cost: 0.0360\n",
      "Epoch: 079/100 | Batch 550/600 | Cost: 0.0288\n",
      "Epoch: 079/100 Train Cost: 0.0555\n",
      "Time elapsed: 16.71 min\n",
      "Epoch: 080/100 | Batch 000/600 | Cost: 0.0642\n",
      "Epoch: 080/100 | Batch 050/600 | Cost: 0.0249\n",
      "Epoch: 080/100 | Batch 100/600 | Cost: 0.0096\n",
      "Epoch: 080/100 | Batch 150/600 | Cost: 0.0314\n",
      "Epoch: 080/100 | Batch 200/600 | Cost: 0.0284\n",
      "Epoch: 080/100 | Batch 250/600 | Cost: 0.0350\n",
      "Epoch: 080/100 | Batch 300/600 | Cost: 0.0299\n",
      "Epoch: 080/100 | Batch 350/600 | Cost: 0.0513\n",
      "Epoch: 080/100 | Batch 400/600 | Cost: 0.0838\n",
      "Epoch: 080/100 | Batch 450/600 | Cost: 0.0479\n",
      "Epoch: 080/100 | Batch 500/600 | Cost: 0.1121\n",
      "Epoch: 080/100 | Batch 550/600 | Cost: 0.0339\n",
      "Epoch: 080/100 Train Cost: 0.0548\n",
      "Time elapsed: 16.92 min\n",
      "Epoch: 081/100 | Batch 000/600 | Cost: 0.0193\n",
      "Epoch: 081/100 | Batch 050/600 | Cost: 0.0571\n",
      "Epoch: 081/100 | Batch 100/600 | Cost: 0.0417\n",
      "Epoch: 081/100 | Batch 150/600 | Cost: 0.0347\n",
      "Epoch: 081/100 | Batch 200/600 | Cost: 0.1274\n",
      "Epoch: 081/100 | Batch 250/600 | Cost: 0.0262\n",
      "Epoch: 081/100 | Batch 300/600 | Cost: 0.0855\n",
      "Epoch: 081/100 | Batch 350/600 | Cost: 0.0893\n",
      "Epoch: 081/100 | Batch 400/600 | Cost: 0.0206\n",
      "Epoch: 081/100 | Batch 450/600 | Cost: 0.0664\n",
      "Epoch: 081/100 | Batch 500/600 | Cost: 0.0120\n",
      "Epoch: 081/100 | Batch 550/600 | Cost: 0.0248\n",
      "Epoch: 081/100 Train Cost: 0.0543\n",
      "Time elapsed: 17.12 min\n",
      "Epoch: 082/100 | Batch 000/600 | Cost: 0.0205\n",
      "Epoch: 082/100 | Batch 050/600 | Cost: 0.0487\n",
      "Epoch: 082/100 | Batch 100/600 | Cost: 0.0307\n",
      "Epoch: 082/100 | Batch 150/600 | Cost: 0.0717\n",
      "Epoch: 082/100 | Batch 200/600 | Cost: 0.0292\n",
      "Epoch: 082/100 | Batch 250/600 | Cost: 0.0810\n",
      "Epoch: 082/100 | Batch 300/600 | Cost: 0.0375\n",
      "Epoch: 082/100 | Batch 350/600 | Cost: 0.0607\n",
      "Epoch: 082/100 | Batch 400/600 | Cost: 0.0556\n",
      "Epoch: 082/100 | Batch 450/600 | Cost: 0.0665\n",
      "Epoch: 082/100 | Batch 500/600 | Cost: 0.0217\n",
      "Epoch: 082/100 | Batch 550/600 | Cost: 0.0703\n",
      "Epoch: 082/100 Train Cost: 0.0533\n",
      "Time elapsed: 17.33 min\n",
      "Epoch: 083/100 | Batch 000/600 | Cost: 0.0655\n",
      "Epoch: 083/100 | Batch 050/600 | Cost: 0.0504\n",
      "Epoch: 083/100 | Batch 100/600 | Cost: 0.0305\n",
      "Epoch: 083/100 | Batch 150/600 | Cost: 0.0655\n",
      "Epoch: 083/100 | Batch 200/600 | Cost: 0.1234\n",
      "Epoch: 083/100 | Batch 250/600 | Cost: 0.0876\n",
      "Epoch: 083/100 | Batch 300/600 | Cost: 0.0826\n",
      "Epoch: 083/100 | Batch 350/600 | Cost: 0.0615\n",
      "Epoch: 083/100 | Batch 400/600 | Cost: 0.0452\n",
      "Epoch: 083/100 | Batch 450/600 | Cost: 0.1022\n",
      "Epoch: 083/100 | Batch 500/600 | Cost: 0.0236\n",
      "Epoch: 083/100 | Batch 550/600 | Cost: 0.0317\n",
      "Epoch: 083/100 Train Cost: 0.0528\n",
      "Time elapsed: 17.54 min\n",
      "Epoch: 084/100 | Batch 000/600 | Cost: 0.0673\n",
      "Epoch: 084/100 | Batch 050/600 | Cost: 0.0207\n",
      "Epoch: 084/100 | Batch 100/600 | Cost: 0.0394\n",
      "Epoch: 084/100 | Batch 150/600 | Cost: 0.0294\n",
      "Epoch: 084/100 | Batch 200/600 | Cost: 0.0215\n",
      "Epoch: 084/100 | Batch 250/600 | Cost: 0.1010\n",
      "Epoch: 084/100 | Batch 300/600 | Cost: 0.0369\n",
      "Epoch: 084/100 | Batch 350/600 | Cost: 0.0703\n",
      "Epoch: 084/100 | Batch 400/600 | Cost: 0.0206\n",
      "Epoch: 084/100 | Batch 450/600 | Cost: 0.0336\n",
      "Epoch: 084/100 | Batch 500/600 | Cost: 0.0471\n",
      "Epoch: 084/100 | Batch 550/600 | Cost: 0.0382\n",
      "Epoch: 084/100 Train Cost: 0.0522\n",
      "Time elapsed: 17.74 min\n",
      "Epoch: 085/100 | Batch 000/600 | Cost: 0.0277\n",
      "Epoch: 085/100 | Batch 050/600 | Cost: 0.0487\n",
      "Epoch: 085/100 | Batch 100/600 | Cost: 0.0470\n",
      "Epoch: 085/100 | Batch 150/600 | Cost: 0.0176\n",
      "Epoch: 085/100 | Batch 200/600 | Cost: 0.0377\n",
      "Epoch: 085/100 | Batch 250/600 | Cost: 0.0619\n",
      "Epoch: 085/100 | Batch 300/600 | Cost: 0.0363\n",
      "Epoch: 085/100 | Batch 350/600 | Cost: 0.0429\n",
      "Epoch: 085/100 | Batch 400/600 | Cost: 0.0681\n",
      "Epoch: 085/100 | Batch 450/600 | Cost: 0.0445\n",
      "Epoch: 085/100 | Batch 500/600 | Cost: 0.0473\n",
      "Epoch: 085/100 | Batch 550/600 | Cost: 0.0406\n",
      "Epoch: 085/100 Train Cost: 0.0513\n",
      "Time elapsed: 17.95 min\n",
      "Epoch: 086/100 | Batch 000/600 | Cost: 0.0467\n",
      "Epoch: 086/100 | Batch 050/600 | Cost: 0.0639\n",
      "Epoch: 086/100 | Batch 100/600 | Cost: 0.0170\n",
      "Epoch: 086/100 | Batch 150/600 | Cost: 0.0407\n",
      "Epoch: 086/100 | Batch 200/600 | Cost: 0.0236\n",
      "Epoch: 086/100 | Batch 250/600 | Cost: 0.0695\n",
      "Epoch: 086/100 | Batch 300/600 | Cost: 0.1013\n",
      "Epoch: 086/100 | Batch 350/600 | Cost: 0.0489\n",
      "Epoch: 086/100 | Batch 400/600 | Cost: 0.0756\n",
      "Epoch: 086/100 | Batch 450/600 | Cost: 0.0649\n",
      "Epoch: 086/100 | Batch 500/600 | Cost: 0.0826\n",
      "Epoch: 086/100 | Batch 550/600 | Cost: 0.0250\n",
      "Epoch: 086/100 Train Cost: 0.0506\n",
      "Time elapsed: 18.16 min\n",
      "Epoch: 087/100 | Batch 000/600 | Cost: 0.0379\n",
      "Epoch: 087/100 | Batch 050/600 | Cost: 0.0714\n",
      "Epoch: 087/100 | Batch 100/600 | Cost: 0.0178\n",
      "Epoch: 087/100 | Batch 150/600 | Cost: 0.0439\n",
      "Epoch: 087/100 | Batch 200/600 | Cost: 0.0323\n",
      "Epoch: 087/100 | Batch 250/600 | Cost: 0.0349\n",
      "Epoch: 087/100 | Batch 300/600 | Cost: 0.0743\n",
      "Epoch: 087/100 | Batch 350/600 | Cost: 0.0183\n",
      "Epoch: 087/100 | Batch 400/600 | Cost: 0.0406\n",
      "Epoch: 087/100 | Batch 450/600 | Cost: 0.0613\n",
      "Epoch: 087/100 | Batch 500/600 | Cost: 0.0188\n",
      "Epoch: 087/100 | Batch 550/600 | Cost: 0.0206\n",
      "Epoch: 087/100 Train Cost: 0.0502\n",
      "Time elapsed: 18.36 min\n",
      "Epoch: 088/100 | Batch 000/600 | Cost: 0.0233\n",
      "Epoch: 088/100 | Batch 050/600 | Cost: 0.0268\n",
      "Epoch: 088/100 | Batch 100/600 | Cost: 0.0329\n",
      "Epoch: 088/100 | Batch 150/600 | Cost: 0.1192\n",
      "Epoch: 088/100 | Batch 200/600 | Cost: 0.0297\n",
      "Epoch: 088/100 | Batch 250/600 | Cost: 0.0315\n",
      "Epoch: 088/100 | Batch 300/600 | Cost: 0.0413\n",
      "Epoch: 088/100 | Batch 350/600 | Cost: 0.0581\n",
      "Epoch: 088/100 | Batch 400/600 | Cost: 0.0321\n",
      "Epoch: 088/100 | Batch 450/600 | Cost: 0.0358\n",
      "Epoch: 088/100 | Batch 500/600 | Cost: 0.0324\n",
      "Epoch: 088/100 | Batch 550/600 | Cost: 0.0780\n",
      "Epoch: 088/100 Train Cost: 0.0496\n",
      "Time elapsed: 18.57 min\n",
      "Epoch: 089/100 | Batch 000/600 | Cost: 0.0165\n",
      "Epoch: 089/100 | Batch 050/600 | Cost: 0.0471\n",
      "Epoch: 089/100 | Batch 100/600 | Cost: 0.0580\n",
      "Epoch: 089/100 | Batch 150/600 | Cost: 0.0141\n",
      "Epoch: 089/100 | Batch 200/600 | Cost: 0.0297\n",
      "Epoch: 089/100 | Batch 250/600 | Cost: 0.0290\n",
      "Epoch: 089/100 | Batch 300/600 | Cost: 0.0317\n",
      "Epoch: 089/100 | Batch 350/600 | Cost: 0.1489\n",
      "Epoch: 089/100 | Batch 400/600 | Cost: 0.0553\n",
      "Epoch: 089/100 | Batch 450/600 | Cost: 0.1766\n",
      "Epoch: 089/100 | Batch 500/600 | Cost: 0.0251\n",
      "Epoch: 089/100 | Batch 550/600 | Cost: 0.0733\n",
      "Epoch: 089/100 Train Cost: 0.0490\n",
      "Time elapsed: 18.78 min\n",
      "Epoch: 090/100 | Batch 000/600 | Cost: 0.0505\n",
      "Epoch: 090/100 | Batch 050/600 | Cost: 0.0507\n",
      "Epoch: 090/100 | Batch 100/600 | Cost: 0.0289\n",
      "Epoch: 090/100 | Batch 150/600 | Cost: 0.0657\n",
      "Epoch: 090/100 | Batch 200/600 | Cost: 0.0292\n",
      "Epoch: 090/100 | Batch 250/600 | Cost: 0.0310\n",
      "Epoch: 090/100 | Batch 300/600 | Cost: 0.0230\n",
      "Epoch: 090/100 | Batch 350/600 | Cost: 0.0274\n",
      "Epoch: 090/100 | Batch 400/600 | Cost: 0.0506\n",
      "Epoch: 090/100 | Batch 450/600 | Cost: 0.0269\n",
      "Epoch: 090/100 | Batch 500/600 | Cost: 0.0521\n",
      "Epoch: 090/100 | Batch 550/600 | Cost: 0.0691\n",
      "Epoch: 090/100 Train Cost: 0.0485\n",
      "Time elapsed: 18.99 min\n",
      "Epoch: 091/100 | Batch 000/600 | Cost: 0.0177\n",
      "Epoch: 091/100 | Batch 050/600 | Cost: 0.0180\n",
      "Epoch: 091/100 | Batch 100/600 | Cost: 0.0331\n",
      "Epoch: 091/100 | Batch 150/600 | Cost: 0.0557\n",
      "Epoch: 091/100 | Batch 200/600 | Cost: 0.0399\n",
      "Epoch: 091/100 | Batch 250/600 | Cost: 0.0463\n",
      "Epoch: 091/100 | Batch 300/600 | Cost: 0.0259\n",
      "Epoch: 091/100 | Batch 350/600 | Cost: 0.1222\n",
      "Epoch: 091/100 | Batch 400/600 | Cost: 0.0707\n",
      "Epoch: 091/100 | Batch 450/600 | Cost: 0.0621\n",
      "Epoch: 091/100 | Batch 500/600 | Cost: 0.0280\n",
      "Epoch: 091/100 | Batch 550/600 | Cost: 0.0276\n",
      "Epoch: 091/100 Train Cost: 0.0480\n",
      "Time elapsed: 19.19 min\n",
      "Epoch: 092/100 | Batch 000/600 | Cost: 0.1132\n",
      "Epoch: 092/100 | Batch 050/600 | Cost: 0.0792\n",
      "Epoch: 092/100 | Batch 100/600 | Cost: 0.0264\n",
      "Epoch: 092/100 | Batch 150/600 | Cost: 0.0432\n",
      "Epoch: 092/100 | Batch 200/600 | Cost: 0.0383\n",
      "Epoch: 092/100 | Batch 250/600 | Cost: 0.0540\n",
      "Epoch: 092/100 | Batch 300/600 | Cost: 0.0595\n",
      "Epoch: 092/100 | Batch 350/600 | Cost: 0.0384\n",
      "Epoch: 092/100 | Batch 400/600 | Cost: 0.0496\n",
      "Epoch: 092/100 | Batch 450/600 | Cost: 0.0225\n",
      "Epoch: 092/100 | Batch 500/600 | Cost: 0.0518\n",
      "Epoch: 092/100 | Batch 550/600 | Cost: 0.0671\n",
      "Epoch: 092/100 Train Cost: 0.0474\n",
      "Time elapsed: 19.40 min\n",
      "Epoch: 093/100 | Batch 000/600 | Cost: 0.0231\n",
      "Epoch: 093/100 | Batch 050/600 | Cost: 0.0628\n",
      "Epoch: 093/100 | Batch 100/600 | Cost: 0.1220\n",
      "Epoch: 093/100 | Batch 150/600 | Cost: 0.0895\n",
      "Epoch: 093/100 | Batch 200/600 | Cost: 0.0408\n",
      "Epoch: 093/100 | Batch 250/600 | Cost: 0.0531\n",
      "Epoch: 093/100 | Batch 300/600 | Cost: 0.0532\n",
      "Epoch: 093/100 | Batch 350/600 | Cost: 0.0866\n",
      "Epoch: 093/100 | Batch 400/600 | Cost: 0.0392\n",
      "Epoch: 093/100 | Batch 450/600 | Cost: 0.0848\n",
      "Epoch: 093/100 | Batch 500/600 | Cost: 0.0314\n",
      "Epoch: 093/100 | Batch 550/600 | Cost: 0.0351\n",
      "Epoch: 093/100 Train Cost: 0.0467\n",
      "Time elapsed: 19.61 min\n",
      "Epoch: 094/100 | Batch 000/600 | Cost: 0.0759\n",
      "Epoch: 094/100 | Batch 050/600 | Cost: 0.0599\n",
      "Epoch: 094/100 | Batch 100/600 | Cost: 0.0419\n",
      "Epoch: 094/100 | Batch 150/600 | Cost: 0.0833\n",
      "Epoch: 094/100 | Batch 200/600 | Cost: 0.0858\n",
      "Epoch: 094/100 | Batch 250/600 | Cost: 0.1197\n",
      "Epoch: 094/100 | Batch 300/600 | Cost: 0.0337\n",
      "Epoch: 094/100 | Batch 350/600 | Cost: 0.0843\n",
      "Epoch: 094/100 | Batch 400/600 | Cost: 0.0424\n",
      "Epoch: 094/100 | Batch 450/600 | Cost: 0.1032\n",
      "Epoch: 094/100 | Batch 500/600 | Cost: 0.0275\n",
      "Epoch: 094/100 | Batch 550/600 | Cost: 0.0312\n",
      "Epoch: 094/100 Train Cost: 0.0462\n",
      "Time elapsed: 19.81 min\n",
      "Epoch: 095/100 | Batch 000/600 | Cost: 0.0388\n",
      "Epoch: 095/100 | Batch 050/600 | Cost: 0.0456\n",
      "Epoch: 095/100 | Batch 100/600 | Cost: 0.0156\n",
      "Epoch: 095/100 | Batch 150/600 | Cost: 0.0167\n",
      "Epoch: 095/100 | Batch 200/600 | Cost: 0.0299\n",
      "Epoch: 095/100 | Batch 250/600 | Cost: 0.0244\n",
      "Epoch: 095/100 | Batch 300/600 | Cost: 0.0254\n",
      "Epoch: 095/100 | Batch 350/600 | Cost: 0.0810\n",
      "Epoch: 095/100 | Batch 400/600 | Cost: 0.0948\n",
      "Epoch: 095/100 | Batch 450/600 | Cost: 0.0414\n",
      "Epoch: 095/100 | Batch 500/600 | Cost: 0.0312\n",
      "Epoch: 095/100 | Batch 550/600 | Cost: 0.0352\n",
      "Epoch: 095/100 Train Cost: 0.0458\n",
      "Time elapsed: 20.02 min\n",
      "Epoch: 096/100 | Batch 000/600 | Cost: 0.0598\n",
      "Epoch: 096/100 | Batch 050/600 | Cost: 0.0425\n",
      "Epoch: 096/100 | Batch 100/600 | Cost: 0.0190\n",
      "Epoch: 096/100 | Batch 150/600 | Cost: 0.0310\n",
      "Epoch: 096/100 | Batch 200/600 | Cost: 0.0615\n",
      "Epoch: 096/100 | Batch 250/600 | Cost: 0.0343\n",
      "Epoch: 096/100 | Batch 300/600 | Cost: 0.1477\n",
      "Epoch: 096/100 | Batch 350/600 | Cost: 0.0405\n",
      "Epoch: 096/100 | Batch 400/600 | Cost: 0.0412\n",
      "Epoch: 096/100 | Batch 450/600 | Cost: 0.0877\n",
      "Epoch: 096/100 | Batch 500/600 | Cost: 0.0237\n",
      "Epoch: 096/100 | Batch 550/600 | Cost: 0.0359\n",
      "Epoch: 096/100 Train Cost: 0.0452\n",
      "Time elapsed: 20.23 min\n",
      "Epoch: 097/100 | Batch 000/600 | Cost: 0.0224\n",
      "Epoch: 097/100 | Batch 050/600 | Cost: 0.0450\n",
      "Epoch: 097/100 | Batch 100/600 | Cost: 0.0584\n",
      "Epoch: 097/100 | Batch 150/600 | Cost: 0.1063\n",
      "Epoch: 097/100 | Batch 200/600 | Cost: 0.0253\n",
      "Epoch: 097/100 | Batch 250/600 | Cost: 0.0496\n",
      "Epoch: 097/100 | Batch 300/600 | Cost: 0.0188\n",
      "Epoch: 097/100 | Batch 350/600 | Cost: 0.0588\n",
      "Epoch: 097/100 | Batch 400/600 | Cost: 0.0442\n",
      "Epoch: 097/100 | Batch 450/600 | Cost: 0.0168\n",
      "Epoch: 097/100 | Batch 500/600 | Cost: 0.0466\n",
      "Epoch: 097/100 | Batch 550/600 | Cost: 0.0362\n",
      "Epoch: 097/100 Train Cost: 0.0446\n",
      "Time elapsed: 20.43 min\n",
      "Epoch: 098/100 | Batch 000/600 | Cost: 0.0504\n",
      "Epoch: 098/100 | Batch 050/600 | Cost: 0.0419\n",
      "Epoch: 098/100 | Batch 100/600 | Cost: 0.0664\n",
      "Epoch: 098/100 | Batch 150/600 | Cost: 0.0394\n",
      "Epoch: 098/100 | Batch 200/600 | Cost: 0.0217\n",
      "Epoch: 098/100 | Batch 250/600 | Cost: 0.0274\n",
      "Epoch: 098/100 | Batch 300/600 | Cost: 0.0186\n",
      "Epoch: 098/100 | Batch 350/600 | Cost: 0.1492\n",
      "Epoch: 098/100 | Batch 400/600 | Cost: 0.0611\n",
      "Epoch: 098/100 | Batch 450/600 | Cost: 0.0307\n",
      "Epoch: 098/100 | Batch 500/600 | Cost: 0.0458\n",
      "Epoch: 098/100 | Batch 550/600 | Cost: 0.0247\n",
      "Epoch: 098/100 Train Cost: 0.0444\n",
      "Time elapsed: 20.64 min\n",
      "Epoch: 099/100 | Batch 000/600 | Cost: 0.0574\n",
      "Epoch: 099/100 | Batch 050/600 | Cost: 0.0363\n",
      "Epoch: 099/100 | Batch 100/600 | Cost: 0.0292\n",
      "Epoch: 099/100 | Batch 150/600 | Cost: 0.0241\n",
      "Epoch: 099/100 | Batch 200/600 | Cost: 0.0201\n",
      "Epoch: 099/100 | Batch 250/600 | Cost: 0.0326\n",
      "Epoch: 099/100 | Batch 300/600 | Cost: 0.0414\n",
      "Epoch: 099/100 | Batch 350/600 | Cost: 0.0445\n",
      "Epoch: 099/100 | Batch 400/600 | Cost: 0.0398\n",
      "Epoch: 099/100 | Batch 450/600 | Cost: 0.0755\n",
      "Epoch: 099/100 | Batch 500/600 | Cost: 0.0270\n",
      "Epoch: 099/100 | Batch 550/600 | Cost: 0.0354\n",
      "Epoch: 099/100 Train Cost: 0.0437\n",
      "Time elapsed: 20.85 min\n",
      "Epoch: 100/100 | Batch 000/600 | Cost: 0.0742\n",
      "Epoch: 100/100 | Batch 050/600 | Cost: 0.0187\n",
      "Epoch: 100/100 | Batch 100/600 | Cost: 0.0329\n",
      "Epoch: 100/100 | Batch 150/600 | Cost: 0.0299\n",
      "Epoch: 100/100 | Batch 200/600 | Cost: 0.1216\n",
      "Epoch: 100/100 | Batch 250/600 | Cost: 0.0615\n",
      "Epoch: 100/100 | Batch 300/600 | Cost: 0.0538\n",
      "Epoch: 100/100 | Batch 350/600 | Cost: 0.0272\n",
      "Epoch: 100/100 | Batch 400/600 | Cost: 0.0597\n",
      "Epoch: 100/100 | Batch 450/600 | Cost: 0.0367\n",
      "Epoch: 100/100 | Batch 500/600 | Cost: 0.0375\n",
      "Epoch: 100/100 | Batch 550/600 | Cost: 0.0281\n",
      "Epoch: 100/100 Train Cost: 0.0434\n",
      "Time elapsed: 21.05 min\n",
      "Total Training Time: 21.05 min\n"
     ]
    }
   ],
   "source": [
    "class MlpSigmoidMSE(torch.nn.Module):\n",
    "\n",
    "    def __init__(self, num_features, num_hidden, num_classes):\n",
    "        super(MlpSigmoidMSE, self).__init__()\n",
    "        \n",
    "        self.num_classes = num_classes\n",
    "        \n",
    "        ### 1st hidden layer\n",
    "        self.linear_1 = torch.nn.Linear(num_features, num_hidden)\n",
    "        self.linear_1.weight.detach().normal_(0.0, 0.1)\n",
    "        self.linear_1.bias.detach().zero_()\n",
    "\n",
    "        ### Output layer\n",
    "        self.linear_out = torch.nn.Linear(num_hidden, num_classes)\n",
    "        self.linear_out.weight.detach().normal_(0.0, 0.1)\n",
    "        self.linear_out.bias.detach().zero_()\n",
    "        \n",
    "    def forward(self, x):\n",
    "        out = self.linear_1(x)\n",
    "        out = torch.sigmoid(out)\n",
    "        logits = self.linear_out(out)\n",
    "        probas = torch.softmax(logits, dim=1)\n",
    "        return logits, probas\n",
    "\n",
    "    \n",
    "#################################\n",
    "### Model Initialization\n",
    "#################################\n",
    "    \n",
    "torch.manual_seed(RANDOM_SEED)\n",
    "model = MlpSigmoidMSE(num_features=28*28,\n",
    "                      num_hidden=100,\n",
    "                      num_classes=10)\n",
    "\n",
    "model = model.to(DEVICE)\n",
    "\n",
    "optimizer = torch.optim.SGD(model.parameters(), lr=0.1)\n",
    "\n",
    "#################################\n",
    "### Training\n",
    "#################################\n",
    "\n",
    "def compute_loss(net, data_loader):\n",
    "    curr_loss = 0.\n",
    "    with torch.no_grad():\n",
    "        for cnt, (features, targets) in enumerate(data_loader):\n",
    "            features = features.view(-1, 28*28).to(DEVICE)\n",
    "            targets = targets.to(DEVICE)\n",
    "            logits, probas = net.forward(features)\n",
    "            loss = F.nll_loss(torch.log(probas), targets)\n",
    "            # or better (more numerically stable):\n",
    "            # loss = F.cross_entropy(logits, targets)\n",
    "            # see \n",
    "            # ../../other/pytorch-lossfunc-cheatsheet.md\n",
    "            curr_loss += loss\n",
    "        return float(curr_loss)/cnt\n",
    "\n",
    "\n",
    "\n",
    "start_time = time.time()\n",
    "minibatch_cost = []\n",
    "epoch_cost = []\n",
    "for epoch in range(NUM_EPOCHS):\n",
    "    model.train()\n",
    "    for batch_idx, (features, targets) in enumerate(train_loader):\n",
    "        \n",
    "        features = features.view(-1, 28*28).to(DEVICE)\n",
    "        targets = targets.to(DEVICE)\n",
    "            \n",
    "        ### FORWARD AND BACK PROP\n",
    "        logits, probas = model(features)\n",
    "        \n",
    "        #y_onehot = to_onehot(targets, model.num_classes).to(DEVICE)\n",
    "        \n",
    "        cost = F.nll_loss(torch.log(probas), targets)\n",
    "        optimizer.zero_grad()\n",
    "        \n",
    "        cost.backward()\n",
    "        minibatch_cost.append(cost)\n",
    "        ### UPDATE MODEL PARAMETERS\n",
    "        optimizer.step()\n",
    "        \n",
    "        ### LOGGING\n",
    "        if not batch_idx % 50:\n",
    "            print ('Epoch: %03d/%03d | Batch %03d/%03d | Cost: %.4f' \n",
    "                   %(epoch+1, NUM_EPOCHS, batch_idx, \n",
    "                     len(train_loader), cost))\n",
    "        \n",
    "    cost = compute_loss(model, train_loader)\n",
    "    epoch_cost.append(cost)\n",
    "    print('Epoch: %03d/%03d Train Cost: %.4f' % (\n",
    "            epoch+1, NUM_EPOCHS, cost))\n",
    "    print('Time elapsed: %.2f min' % ((time.time() - start_time)/60))\n",
    "    \n",
    "print('Total Training Time: %.2f min' % ((time.time() - start_time)/60))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(range(len(minibatch_cost)), minibatch_cost)\n",
    "plt.ylabel('Cross Entropy')\n",
    "plt.xlabel('Minibatch')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(range(len(epoch_cost)), epoch_cost)\n",
    "plt.ylabel('Cross Entropy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuracy: 99.00\n",
      "Test Accuracy: 97.67\n"
     ]
    }
   ],
   "source": [
    "def compute_accuracy(net, data_loader):\n",
    "    correct_pred, num_examples = 0, 0\n",
    "    with torch.no_grad():\n",
    "        for features, targets in data_loader:\n",
    "            features = features.view(-1, 28*28).to(DEVICE)\n",
    "            targets = targets.to(DEVICE)\n",
    "            a1, a2 = net.forward(features)\n",
    "            predicted_labels = torch.argmax(a2, 1)\n",
    "            num_examples += targets.size(0)\n",
    "            correct_pred += (predicted_labels == targets).sum()\n",
    "        return correct_pred.float()/num_examples * 100\n",
    "    \n",
    "print('Training Accuracy: %.2f' % compute_accuracy(model, train_loader))\n",
    "print('Test Accuracy: %.2f' % compute_accuracy(model, test_loader))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  },
  "toc": {
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}