R2_square_neural_network_cup_net.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "from torch.nn import Linear\n",
    "from torch import optim\n",
    "import os\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from torch.utils.data.dataloader import DataLoader\n",
    "from torch.utils.data import random_split\n",
    "import matplotlib.pyplot as plt\n",
    "import torch.nn.functional as F\n",
    "from torchvision.transforms import ToTensor\n",
    "from torchvision.utils import make_grid\n",
    "from torch.utils.data.dataloader import DataLoader\n",
    "from torch.utils.data import random_split\n",
    "from mpl_toolkits import mplot3d\n",
    "%matplotlib inline\n",
    "torch.set_printoptions(linewidth=150)\n",
    "import math"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Plotting the data for a single cup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "size of dataset is  900963\n",
      "the size of geometric points for each cup is  2043\n"
     ]
    }
   ],
   "source": [
    "dataset = pd.read_csv('D:/JOB, Application STUFF/Fraunhofer/Frauhofer task/Takk 3/Example_Data.csv', sep=';', decimal=',')\n",
    "n=len(dataset)\n",
    "print(\"size of dataset is \", n)\n",
    "m=2043\n",
    "a=int(n/m)\n",
    "print(\"the size of geometric points for each cup is \",m)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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"
    }
   ],
   "source": [
    "fig = plt.figure()\n",
    "x = dataset.loc[0:2043,'X']\n",
    "#print(x.shape)\n",
    "y = dataset.loc[0:2043,'Y']\n",
    "z = dataset.loc[0:2043,'Z']\n",
    "ax = fig.add_subplot(projection='3d')\n",
    "plt.xlabel(\"X coordinate\")\n",
    "plt.ylabel(\"Y coordinate\")\n",
    "#print(z.shape)\n",
    "ax.scatter3D(x, y, z, c=z, cmap='nipy_spectral')\n",
    "plt.show() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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"
    }
   ],
   "source": [
    "fig = plt.figure()\n",
    "x = dataset.loc[0:2043,'X']\n",
    "#print(x.shape)\n",
    "y = dataset.loc[0:2043,'Y']\n",
    "z = dataset.loc[0:2043,'Z']\n",
    "plt.xlabel(\"X coordinate\")\n",
    "plt.ylabel(\"Y coordinate\")\n",
    "xi = np.array(x)\n",
    "yi= np.array(y)\n",
    "plt.scatter(xi,yi)\n",
    "plt.show() \n",
    "#print(xi)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Calculate the hidden size for Refnet by Cupnet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "/9j/4AAQSkZJRgABAQEAeAB4AAD/4RDuRXhpZgAATU0AKgAAAAgABAE7AAIAAAAMAAAISodpAAQAAAABAAAIVpydAAEAAAAYAAAQzuocAAcAAAgMAAAAPgAAAAAc6gAAAAgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHNuZWhhIHZlcm1hAAAFkAMAAgAAABQAABCkkAQAAgAAABQAABC4kpEAAgAAAAMyNAAAkpIAAgAAAAMyNAAA6hwABwAACAwAAAiYAAAAABzqAAAACAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAMjAyMDoxMToxNyAwMjowMDoyMwAyMDIwOjExOjE3IDAyOjAwOjIzAAAAcwBuAGUAaABhACAAdgBlAHIAbQBhAAAA/+ELHmh0dHA6Ly9ucy5hZG9iZS5jb20veGFwLzEuMC8APD94cGFja2V0IGJlZ2luPSfvu78nIGlkPSdXNU0wTXBDZWhpSHpyZVN6TlRjemtjOWQnPz4NCjx4OnhtcG1ldGEgeG1sbnM6eD0iYWRvYmU6bnM6bWV0YS8iPjxyZGY6UkRGIHhtbG5zOnJkZj0iaHR0cDovL3d3dy53My5vcmcvMTk5OS8wMi8yMi1yZGYtc3ludGF4LW5zIyI+PHJkZjpEZXNjcmlwdGlvbiByZGY6YWJvdXQ9InV1aWQ6ZmFmNWJkZDUtYmEzZC0xMWRhLWFkMzEtZDMzZDc1MTgyZjFiIiB4bWxuczpkYz0iaHR0cDovL3B1cmwub3JnL2RjL2VsZW1lbnRzLzEuMS8iLz48cmRmOkRlc2NyaXB0aW9uIHJkZjphYm91dD0idXVpZDpmYWY1YmRkNS1iYTNkLTExZGEtYWQzMS1kMzNkNzUxODJmMWIiIHhtbG5zOnhtcD0iaHR0cDovL25zLmFkb2JlLmNvbS94YXAvMS4wLyI+PHhtcDpDcmVhdGVEYXRlPjIwMjAtMTEtMTdUMDI6MDA6MjMuMjQxPC94bXA6Q3JlYXRlRGF0ZT48L3JkZjpEZXNjcmlwdGlvbj48cmRmOkRlc2NyaXB0aW9uIHJkZjphYm91dD0idXVpZDpmYWY1YmRkNS1iYTNkLTExZGEtYWQzMS1kMzNkNzUxODJmMWIiIHhtbG5zOmRjPSJodHRwOi8vcHVybC5vcmcvZGMvZWxlbWVudHMvMS4xLyI+PGRjOmNyZWF0b3I+PHJkZjpTZXEgeG1sbnM6cmRmPSJodHRwOi8vd3d3LnczLm9yZy8xOTk5LzAyLzIyLXJkZi1zeW50YXgtbnMjIj48cmRmOmxpPnNuZWhhIHZlcm1hPC9yZGY6bGk+PC9yZGY6U2VxPg0KCQkJPC9kYzpjcmVhdG9yPjwvcmRmOkRlc2NyaXB0aW9uPjwvcmRmOlJERj48L3g6eG1wbWV0YT4NCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAKICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgIAogICAgICAgICAgICAgICAgICAgICAgICAgICAgPD94cGFja2V0IGVuZD0ndyc/Pv/bAEMABwUFBgUEBwYFBggHBwgKEQsKCQkKFQ8QDBEYFRoZGBUYFxseJyEbHSUdFxgiLiIlKCkrLCsaIC8zLyoyJyorKv/bAEMBBwgICgkKFAsLFCocGBwqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKioqKv/AABEIAdQB8wMBIgACEQEDEQH/xAAfAAABBQEBAQEBAQAAAAAAAAAAAQIDBAUGBwgJCgv/xAC1EAACAQMDAgQDBQUEBAAAAX0BAgMABBEFEiExQQYTUWEHInEUMoGRoQgjQrHBFVLR8CQzYnKCCQoWFxgZGiUmJygpKjQ1Njc4OTpDREVGR0hJSlNUVVZXWFlaY2RlZmdoaWpzdHV2d3h5eoOEhYaHiImKkpOUlZaXmJmaoqOkpaanqKmqsrO0tba3uLm6wsPExcbHyMnK0tPU1dbX2Nna4eLj5OXm5+jp6vHy8/T19vf4+fr/xAAfAQADAQEBAQEBAQEBAAAAAAAAAQIDBAUGBwgJCgv/xAC1EQACAQIEBAMEBwUEBAABAncAAQIDEQQFITEGEkFRB2FxEyIygQgUQpGhscEJIzNS8BVictEKFiQ04SXxFxgZGiYnKCkqNTY3ODk6Q0RFRkdISUpTVFVWV1hZWmNkZWZnaGlqc3R1dnd4eXqCg4SFhoeIiYqSk5SVlpeYmZqio6Slpqeoqaqys7S1tre4ubrCw8TFxsfIycrS09TV1tfY2dri4+Tl5ufo6ery8/T19vf4+fr/2gAMAwEAAhEDEQA/ACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACisbxait4T1F2UFo4GdGxyrAdR6GqkMGnX9vDYw2vlXP2UTLMYWjKsNoBzgZ5P6e9AHSUUUUAFFczr6QL4q0JpoQ6yNOJVEZfeBHxkAHOD+VX9Ngs7i6XUdPhFuF823dQhTeAwGSuBzlfTvQBr0UUUAFFFFABRRVHWo0k0O93qG2wOwyOhCnB+tNK7sBeorltMi06/0PTrGS2/0i5sQzTGEo2Qi5YPgZOT1Brp0DCNQ5y2Bkj1qpx5XYQ6iiioGFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUVXvr+1021a4vplhiXqzHr7D1Nc1aeJdZ1y/lTRLG1ht1iWVHv2dWlViwDAKDgfKev170Ds7XNzX7O41HQ7qytBH5lxG0eZXKhcjrwDU9jDJHawi6jjWaKMR5RiwxgZ5IHp+lZe7xd/zy0T/v5L/8TRu8Xf8APLRP+/kv/wATQI3qK5XU9V8TaVaefcx6MdzBI443lLyseiqMck0aVqfirVtNjvIbfSESQsAsjyhhhipzgEdR60AaOp6ffXOv6Ze2yW5isTIWEkpVn3pt4wp6VrogRflULk5IHqetYe7xd/zy0T/v5L/8TRu8Xf8APLRP+/kv/wATQBvUVxWpeJfEWmXiWzwaVNKSnmCFpG8kMwVS2QMZJ6dfatnd4r/55aN/38l/+Jq4x5uoG5RWHu8V/wDPLRv+/kv/AMTRu8V/88tG/wC/kv8A8TVez8194rm5VXUoZrjTLiC2VDJLGyDexUDIIzkA1m7vFf8Azy0b/v5L/wDE0bvFf/PLRv8Av5L/APE0KDTvdfeBd0e0mtNKs7a8SLzbWFYg0blgcKATyBjOKv1h7vFf/PLRv+/kv/xNG7xX/wA8tG/7+S//ABNNwbd7r7wNyisPd4r/AOeWjf8AfyX/AOJo3eK/+eWjf9/Jf/iaXs/NfeFzcorD3eK/+eWjf9/Jf/iay9W8ReINHIWeDS5pNpkaOB5CyoOrHIAA7e56VcMPKpLli036hex2FFFFc4wooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAqpLqVrDdi1kZ/OYFlQROcgdSMDmrdYd+ZB4wsPJRXb7LLw7FR1Xvg0m7GkIqTszUhvoLid4Y2YSooZkeNlOD35AqxVSyEssST3key4G9CoPAG7p79BzUl5e29hbNcXkywxL1Zj+nufajoS1rZE9Y+oa8sUz2umoLu7QZk+bEcA9ZG6D6darvLf63Gzln0rSwCWkY7ZpV9f9hffrWTbfZ9XvE060t/s+hxx+cqYwb0hsBm7lM+vXGTxS1exfLGHxavt/mUrO0bxJrvm6rI9zbRRiVCw2pNkkDavaPg9eW6ng4rpdPK/wDCX3irj5bGAEDt88lZ91czL4ne005FM7WaAuR8kK72+Y/0Hf6c1Z0OzjsfFF5HGWdmsoWkkc5aRjJJlifX/wDVTSsRKTk7s6Wqep6nb6VZ+fcbmLEJHEgy8rnoqjuTRqep2+lWfn3G5ixCRxIMvK56Ko7k1haOkupStrGqD/S98kMcOcraqrlSq+pO3lu/0pkho6S6lIdZ1QD7UWeOKHOVtVDFSq+pOOW79OlXPBjK/hO1ZCGUvKQQev716yNJWTWLExKSmnCeYu6nm5zKx2gj+D1Pfp61seDQF8KWqqAAHlAA7fvXoA3KwPEXiB7BWs9MUS3xC7mIyturHAZvfngd/pSeI/EElhHJa6WolvtgZyRlbdScBm9/Qd/pVDVLa20Xw1MdzuzSxvLK3zSTOXXJPqTQAanYQ6b4fMaMXkkuoGklkOXlczJlie5rr64rV7Wa408X+ortkFxB5EGciAGZOT6uR1PboO5Pa0AFFFFABRRRQAUUUUAFFFY+v67/AGVbPHaR/aL3ymkWLsigcu3oP5nge106cqklGO4m7CeINe/sm2dLWP7Re+U0ixdkUDl29B/M8D2ydXso7DwvqLyymW4mTdNcScNI3QfQdgO1PvrSOw8M6lLNL5txPA7TXEnDSMVwPoOwHaq+uQy3uhXl5eqY0SImC3P8P+03+17dvrXoUVFOCi9ObV99vw/4d9ESztKKKK8wsKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKozaWs2qR332mdJY0KKq7NoBxnqvtVi6u4LK3ae7lWKJerMcCsfztR1/i28zTtOPWYjE0w/2R/CPfrUuxrCMt9kWb7WxFcmy02E3t93jQ/LF7u3b6dajttHCzf2jrtwt1cxgsC3EMA6/KD/AOhHmr1pZWek2ZS2RIIUBZ2J6+rMT1+prm5r5vE2rRWzIyaP5bSoDwbsqyjJH/PP5sgd8ZPFFu4OaStD/ghLfN4n1ZLZkZNHEZlUHg3ZDAAkf3OeB3xk8Ut5czDxR9k05VNw1mAWI+WFd/3j6+w7+w5ovrmUeJ47TTlQ3BsyCx+7ApcfMR36cDv7DmkRbXRNWd5HY5tQ0kjfM8rl/wBSegA9gKoyCJLXRNWneR2ObVGkkb5nlcu35k9AB7AU221ZtO12+u9VTY8lpAILaMbpCS8m1B/eY8njgfQZqtJcC28QPqWsR/vvsyC1tUG5gSzAKB0LnnkdMntzVzQ7V7y8n1fVI0F8HaBFHK26KT8oPc8nJ9+OKAJNHSbUZ5NW1UD7WskkMcQOUtlVipC+pOOW7/SqulpLq1rLCMppwurgu6tzc5mc4BH8GDye/T1o0tZNYiuIlJTThdzl3U83OZGO0Efwep79PWpbC8NtYGzsI1e6e6ufLj6LGvnuNzY6KP16CgA0+8NtpqWdhGr3TzT+XH0WNfNcbmx0Ufr0FUNP1+bTfBcVpp2LnUUWZ5HI+WBfNf529z2Xv9Khg1B9J0O5i01VudU3Ty3M7DAQB3G9vTOOF/pmtLUbGy8P+DruOLgFCXdvvSue59SaAH6rbW2i+F7rLM7MQ0krnLzOWHJ9SabqVpNc2DX2ortkDx+RBnIgBdeT6ue57dB3JTU7Sa50ya+1FdsgA8iDORACw5Pq57nt0HclNfuH1Owmt7Q4tI5EWecH753gbEPt3Pbp1zgANdvZLy1KWYU2sNzCJpiOHbzkGxfoep9seuOzrkvEMttZ6Tb2UYVC88AiiReirKnOB0A9fcetdbQAUUUUAFFFFABRRWRrmtHTozBZRie+aMukZ6Io6u/oo/U8CrhCVSXLENhNd1v+zIGis4xcXxjaRY+yKOrv6KP1PArLubaOy8MahPNL5lxcW7vPPJ96RipA+g5wAOlK8MVl4ZvLq4l3z3Fu0k88n3pGK8fQc4AHSoLqCW+0C4u75CiJaOYLdv4fkPzt/te3b613wjFJJbX1ff8A4H/DvoiBdRglvtDu7u+QoiWzmC3b+D5T87f7Xt2+tHiW7kuNIvreyxsiQ+fMRkD/AGB6t6+n1p+t3UtzpF9b2JGyK3fz5iMgfL9werHv6fWl8Rtbad4XntoxsDRlIo1GSfU/1JqqV+anda30Xbb+vxYPqdbRRRXllhRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUVFcXMNpbtPcyrFEgyzucAUBuS1l3+tpb3H2Oxia9vyOIYzwnu7dFFVftOoa/wAWG/T9PPW5YYllH+wD90e5rUsNNtdMt/Js4ggJyzdWc+pPUmpu3sa8sYfFv2/zKFrojzXC3uuSrd3K8pEB+5h/3V7n3PNa7uscbPIwRFBLMxwAPU0O6xxs8jBEUEszHAA9TXI3F8/iTVre2ZGTRnV5FB4N2UK8kf8APP5uB3xk8U0rESk5biz37eJdWgtWRk0do3mQHg3ZRkGSP+efz5A74yeKW/uZf+Emt7TTvLNz9jkDE9IVLp8xHfpwO/sOaXUbiX/hJbS207yzci0mDbjxCpeLDEd/unA7nHbmmstromrQvI7HNrK0kjfM8rl4/wAyegA9gKZIBbXRNWDyOxzas0kjfM8rl1/MnoAPYCqk1wLbxANS1iP979mAtbVBuZSWICj1c88jpz2GaLi4+y+IItS1iP8AeG2YWtog3MrFgAo9XPOccD6DNXNEt5bu/utT1ZE+2xyGCNFOVt02g7Qe5+Y5bv24oANDtpLy9uNV1WNPtyOYEQHK26D+EHueTk/lxUGmB9ZF3Eh26cLuXe6nm5+b7oP9z1Pfp60abu1p76ONsaaLuTe6nm4PHyg/3fU9+nrU1neG2jns7CNXunupfLj6LGu7G5sdFH69BQAWN4ba2ks7CNXunup/Lj6LGvmsNzY6KP16CsuLUX0rR76PTUW51MSXM1zMRgIFkf529zjhf6ZpYL19L03UIdMVbnVDLcTXE7DAQK7fO35cL/8AXrRvrW00fwXfjcA01vI0krn5ppHU8k9ySaADUrOz0PwdforY3wuXlc/NLIw6k9yTSapaTXOkXV9qK7ZBEfIgzkQA9z6ue57dB3JNTtJrnQ72+1Fdsi2z+RBnIgBU8n1c9z26DuSniG4k1PSb2CzbbaxIRNOP4yP4F/qfw65wAHiC4k1PTbqC0O20jIWecfxncPkT+p/DrnE3iO4h0/RFtLWNfNZo1ggUYGA68nHRR6/SjxHcQ6foLWltGPNZQsMCDGACOfZR6/40mpWiWGiXN1eSiS4do2mnbgYDrwPRR2H9aAG6pY/ZdHkmnfzrua4t/NlIxn98mFA7KOw/rXYVxOsG41DT1vJN8FqlxAYITw0mZUG9x9Dwv4nnp21ABRRRQAUUVlaxq7WWLWxRZ7+RSyRn7sa/339FH5k8CrhCU5cqDYTWtZNgv2eyQT38iFkjJ+VFHV39FH5k8CsuOOKz8N3F5cy7pri3MtxPJ1div8ucADpRbCGz8OS391LumuYPOuJ5OrsV6fQZwAKhgtpbzQ1ur5dqR2uYIP7vyffb/a9B2+td0YxjGy2T1ffy/r1fRECG3lvdBe7vl2oloTBbn+H5Pvt/te3b607VLqW50S6t7EgJFat58xGQvyfcHqx7+n1pLu5lu9DltrE4SO1Pnz9l+T7i+revp9al1J7bTfCb28a7d1qyRRoMknYcn+pNWr80brW+i7f1/wAFgLrL22m+F57eNdoa3dIo0GSflOT/AFJqLWLNovD2oXN2wku5ISGYfdQf3V9B/PrT7yzaLw/qFzdsJLqS0cMw+6g2n5V9B/PrVbXnk1PQbyYFo7KOMtGOhnPZj6L6evXp1KO8bP7Wr+7+vxBna0UUV5ZYUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUVRvr5onFtaAPcsM8/djH95v6DvSbsVGLk7Ij1HWUs5ltLaF7u+cZS3j7D1Y9FHuar2+iS3Vwt5r8i3UynMduo/cw/QfxH3NN0OHydW1MFmkb90WdzkscHJrcqV72rNqlqT5Ydlr6q/yCmu6xxs8jBEUEszHAA9TQ7rHGzyMERQSzMcAD1Nchd30niLU7SEoU0WRmIVuDdlRnJH9zOMDv1PGKs5xbq+k8R6pawFGTRn3uoPBuymOSP7mTwO+MnjFO1O5k/4SOxtNO8s3P2eYNuPEKkx/MR36HA7/TmjVbmT/hINPs9O8s3PlS5LfdhU7fmI79Dgd6SZbXRNTs3kdjmCdpJG+Z5XLRfmT0AHsBQASLa6Jq9q8jsc2tw0kjfM8rl4fzJ6AD2Aqpc3Jtdft9R1hPnNtILW0RQzKxZAFHq5yc44H0GaW7umtdds9Q1dPna3m+y2iKGZW3RAKPVyCc9h9ASbeiwT3Wp3l9q6L9sgkEMaK2VgQxo+B7/Pgt3x6UAGiQTXeo3eo6sifbYZPJjRTlYE2K20e/zct3+lQadu1q41BInxpv2tt7qebg7VG0f7PHJ79BxmlsA2tXmpLFIBppu/ndDzORGg2j0Xjk988cZqW2vDbTXtnYRq90903lx9FjXYo3Njoo/XoKAC0vDbG6s7CNXunun8uPosa8Dc2Oij9egrNtruTTLXUbfTQLnVHmmluLhxwgBPzt+Xyr/9eltbmXTo9StNN/0nU5J5JZ7iQcRqB99v12r/APXrQ8u10nwZPK7Yaa3MksjH5ppHXqfUkmgBJ4LTSPBd27MA01s7yyufmmkZepPckmmX1pLceHL2+1FcSLZSeRBnIhHlnn3c9z26DuSSWktx4dnvtRA8xbJ/IgByIR5Z592Pc9ug7kprF1LqHh++isW2W0NrJ504Gd5CH92v9T+A5zgAdr9zLqOjX8Vk2y1hhfzpwM7yB9xf6n8BznE3iS4h0/w7LaW0YMjxFYYIx0A6n2Uev+NHiO4g0/w1PaQR/O9syRQxjooXk+yj1/DqaTUrMWXh7ULi6kEtzJETNMRj6KB2Udh/U0AGpWi2Og3lzdyiS5kCmacjAwGHA9FHYf1NQ6skuq6c11cB4rNJIzDAeDJ86/O49PRfxPPANXSXVdKmublWjs02mCBuDJ8w+dx6ei/ieeBJ4lvHms5bGw5kDR+dN2gBdcfVj6enJ7ZAF8R3weD7Jbp5pS4gM7g/LEPNTA92PHHpz6Z6+uS1qC207QI7eIhA11AF3NlpG85SST3PBJrraACiiszVdVa0dLOxRZ9QmGUjJ+WNf77+ij8yeBVwg5uyATVtWa0ZbSxRZ7+ZcpGT8sa/339FH5k8CsrTzDa+HjqN1JmW6gE9xPIclmK5x9BnAAp2lGG10Qajdy5muIxNcTydXYj+Q6ACq+kWstzotncX2NkVsnkQg5Awg+c+pPb0+tdvLGMHHomrvvvp/Xq+xIlhay3eiQXN8MJHajyIM5A+T77erenp9aHuJbzQvsti21I7UCecfw/J9xf9r+X1pI7iW90GK0sWKIloonuB/B8g+Rf9r37fWpnmh07wnFFGmWe12xRIOXOzJ/xJrR35tVrfRdv6/wCCxBdS2+m+ExDGmC9qUiiQcsdnP+JNNls2i8OXlzdsJLqSycMw+6i7D8q+g/n1oS0MPhy4urtxJcvZEM/QIuz7q+g/n1qG9Mmq+H7iU7o7KO1Zox0M5CcMfRfQd+vTqo76P7Wr/r+uoC6oZNV0K7my0djHbu0Y6GchThj6L6Dv16dZPFN5jR7u0t08yXyt0npEvqfc9h/Sn6/eEaPd2lqgkl+ysZOcCJNp5Puew7/Sm65BBp3hO9Xfy6HdI5y0jnuT3JopW5qd19rRfdr/AFv6A+p1tFFFeYWFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUVQ1HUGgzBaAPcld3PSMf3m/oO9JuxUYubshb6+aJxbWgD3LDPP3Yx/eb+g71Q0tNtvKXYvIZ5N8jdWwxGT+Ap+nIBYxysd0kyCSRz1ZiOSag021k1COQSApZedIxIPM+XJAH+z/P6VHVM6uVckoLRJq7+/8ApIjstVt7PWLmS63xW95sWG5ZcRsVBBG7t7djXRtIiRtI7KqKNxYnAA9c1HcRW7Wbx3SRm3CfOrgbQo9e2K4aOF72+tbeCWb/AIR2aciK3mOfO2ozZHcR5UYB69apJoxqSjPXZ6fhoXr2+l8RalZxbSmiySMArcG72qWDEf3MgYHfqeMU/V7qT+3tOs9O2G62yElvuwqV+8R+eB3xRrN0/wDbWm2WnlPtW5ySwysKlCNxx+OB3xSXS2uiXlg8jscmVpJG+Z5XKgfiT0AHsBVGIXC2uiahYvI7HKTNJI3zPK52D8SegA9gKq3ty1rrdjqOroS5hmFtaIoZkYmMAD1c5Oew/Ami+na21mw1PVkYvslFtaIoZkJ2gAernPPYfhmrejwXFxrF3eauEN3DtSKNTlbdWXJUep55bv8ASgA0aGe51W8vdXVftcLCOKNTlbdGRWIHvzy3fHpUNkrazqGqLFIP7NN0DI6HmciKMbPZeOT3zxxmi0Da1quppFIBpvnr5joeZyI1G0ei8cnv24zUsN4bbUdTs7CNXunuV8uPosa+RENzY6KP16CgAgvDbXmoWdhGr3T3X7uPosa+VGNzY6KP16CqFpPNYPqVjpx+06nJcNJNcyLxGu0fO36hV/pmnWk09nd6lp+nsbjUpbjfLcyrxEvlp87fjkKv9KvaebbTfCCXUpw9xAJppDy0sjrn6kknAH4UAJbi20vwc1xIcPPb+bLIeWmkdf1JJ4FRw2cs/h43uoAb1sj5EIOVhHl9fdiOp7dB3JNPs5Z/D8V7qGC62QEEI5WIeX192I79ug75jlnl1Tw48Fm5S1jtMT3A/jITlE/qfwHPQAW/uJdS8O3MNk+y1itGE1wB98hOUX+p/Ac5xPrlxBYeFJrSGPLyWTpHDGOihOW9lA7/AIdTRqlzFYeEjawR7pZLJlihjHQBOWPoo7n+pou7IWnhPUp55PNuprJ/NlIx/AcKPRR2H9TQAuo2Qs/DWpz3Mnm3Uts5lmIxn5ThQOyjsP6moNZjk1bRrq5uVaOzSMtBAwwZD2dx/JfxPPANYjk1bQby5uVaOzjgZoIGGDIQOHcfyX8TzgCXxNeyS6ZeWVgA0ix5nlP3YR6e7HsPxPbIAeJrySbT7mysMGRQvny/wwjIx9WPYfie2X62LXR/DbRxg4MibVzl5XLgnr1Y8mjXfsuj+F5o0yASAq5y8rlgfqWPWmahZyHT5L7UdrXTNGFQHKwL5i/Kvv6nv9KAGatZSnTTfahta6aeAIgOVgUzJ8q+/qe/0rs64rWrubULISWpC2MVzADJj/j4bzkGF/2R69yOOOvRarqrWjpZ2KLPqEwykZPyxr/ff0UfmTwKuEHN2QBquqtaOlnYos+oTDKRk/LGv99/RR+ZPArL0gxWmjf2heS7ppx5txcP1c9vwHQAfhRpBitNIa/vJd00xL3Fw/Vzkgfh2AH4VBoFtJd6faXN8MJGv7iDsv8Att6t6en1rscYxpyj0TSb776en/DvoieoaFay3Wl2VxfABIol8iHOQOPvt6t6en1pllPLe+H7W0sXKIlognuF/g+QfKv+179vrTtMnlvdHtLSxcoiQIJ7hf4PlHyr/te/b61JaXMVh4TsVSPfLJaoI4U4MjFMn/EmtZ83NK61vou2/wDX4sQJcRaf4TtUjj3SSWoEUKdXYpk/4k0lpaCDw693dSCSd7P5nPAjXZ90eg/n1o020WDw8t3cyb5msxl24EabMhR6D19TzUKLJqmgK0gaOyjtflQ8Gchep9F9B3+nVaXdnpzav9F/XqAkok1Xw+8jbo7KO1JRehnITqfRfQd/p1n1i8I0Oe0tUEk32NjJz8sSbDyfc9h3+lJf3jDQWtLNQ85s8uf4Yl2dT7+g7/SnXkdvpnhC6BfG+2bc7n5pHK45PcmhbxuvtaL7tf639AF1KK30zwneqXwXgfdI5+aRyuMk9yar61bSz6Fe3t+MSCE+TDnIhB/mx7n8B7vv7aWfQ7y9v1xILWTyYc5EI2H82Pc9ug92eIrmS90a+S0O22iQ+bN/fI/gX+p/D1w6N+aNnrzav7tP6/IGdlRRRXllhRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUVn6jqQt5Y7S3KtdzHCg8hBgnc34A4HfFJuxcIObshurap9ijMVsBJcEDjtGCcbm/oO9Qrbrb20mCXdgWeRurnHU1Xu7dbfTZOS7s6s8jdXO4cmp4ojqzd1sQeSP+W/sP9n37/TrGt9TptHkXLtfV99v6SINMgfU7G3VwyWaRqG7GYgdP93+f067zMkMJZiscaLkk8BQP5Cqt/qVppNurXD7c/LHEgyzn0VR1rl9fi1PWI7KPUXawsbu6EX2WIjzGXYzZc9Oqj5elNaaGdRufvbL+vvH399L4gvrFNu3RZrgxhWBDXeI3cMfRMoMD+LqeMU/Wrpv7W0yy08p9r81m+YErEvlOMnH44HfFZOraS8eoafY2Otai9003/LWbKwr5b/NwvBwDgdxn607UdCt9Jms5X1fUwJbht7+cCxby2wRhcljgDv1xVnOad4trolxpzyOxzLI0kjfM8rmMj8SegA9gKq6hO9tqun6nqqNv/eLb2kY3MhKgAD1c557D8CTV/wCEeuZL/T5L/U9RV5JJPLUzgtCuwnrjG7jnH096LLR4tS8Q3wGs6lILBUWKYTgkFt28A7f9nHHvQBqaRBcz63d3esBDdRIghjU5W2VgSVHqemW7/SobXOt6xqSRPjTt6eY6nBuCEA2g/wB3jk9+g4zWZZaM2raxfxw6vqRsECq8pmG64JB4BxgqOee/b1px0t4NTu7Cw1TUmusoI1WcAKuwfM+F+6Og9egoA147w22qajZ2EavdPKnlx9FjXykG5sdFH69BVS0luLXUdT06wY3GpS3CtJcyrxEhij+dvxLbV/oKhttCjtLjVJrrW9RiMToJZ1mCmT92p54JPJwB9BTdJ0uGPw7Fqdxq2pQvcgSS+VMMyOeOBtySeg6mqik93YDV0l7bTfCkN5McPcRLNM5+Z5ZGAP1JJOAPwqLRbKSbQbS8vyGZLNRbxD7sS7MbvdiO/boO5NDRPDkt3odnd3upXqsIg8EcUw2xKV4xkdcfkDj3MGnadLqWkwfZdU1BbVLUee3nfIW2f6tBjoO57dOTnF8sP5hGjFPJqnhtLezcpaR2YE9wv8ZCcon9T26DnpYu7qOx8JR20Ee+aSy2xQp6bOWPoB3P9TWVJpwtPDdsINS1J55rUGK1jnAX7mTkY4QZ5/LqakXw8sHhx76fVb4TGy3SETAKQEzt6Z2+2aOWH8wGg1mtn4TvLi4k824lsmMszemw4UeijsP6moNSjl1Xw3eXNyrR2kdm7QQHgyEIcO/8wv4nnGM9tEmv/DtxdXV/qH2RbYyQQSyg78LncwxwPQde+ewXWLGUaXdQ2OqajO0Ns0lx5s4MaLsztIA5JHb05PbJyw/mA1vEt7JLpF9Z2IDOtuxuJD92JdvT3YjoO3U9sv10Wmj+FLmJflDKVUZy0rn9ST1JrM1rRbfR/Dt0/wDauogMjbYjMCJXI6EBcnPf8TSX/hyZNFmvdQ1K9e5SNiEMwZUU/wAOcc+5GM0csP5gNHUbOQ6Xc32o7WumUBUBysC7h8q+/qe/0qLXJX1nTZVhO3TldFd+9yd4GF/2ffv2465uoaZJe2NzJDqmoS6eg2lppgROdwHyjH3R69+3rUniDTfsVh9ms9U1Ga5bZ5cLzAoi7gNzYXhR29+lHLD+YDY15/Nt49J09Va7kaN0QD5YkR1bc3ovy49+1P0jy7PSZL69m3zSOzXFw/VyGIH4YGAB+FR6AIbHQ3u7qXdI0knn3Ehy0m12UZ/AAAU3Qbd7yzhu7wERo7tBAw+6d5+Zv9r+X1rqcYwpyj0TSb776f16voieoaDbSXljbXV6MJGSYID/AA8n52/2vT0+tJpk8t7pdvaWLlERAJ7hf4P9lf8Aa9+31pdNnlvdPjs7JiipkT3A/g5Pyr/te/b61JplxFp3h2ySKPdI6YihTq7df/rk1pUveWmt9F23/r8WCF065i0/w1YLHHvkkhURQpwZGIyf8SabotokHh+C7nk3ytZp87cCNNgwo9B6+p5pfD1osGjW93cSb5mt1y7cCNMcKPQevqear6bFJqeg2guFaOyitkAjPBnYKOT/ALOe3f6dVK3vpPTm1f37f16gFrHJqmgQ+crR2UdsuEPBnIXqf9n0Hf6dZZ7t/wCwUs7IBpzZguf4YV2dT7+g7/SkF2/9gW9nZANcGyUsx+7Cuz7x9/Qd/pT0W20vwgWJxutssxOWkcp+pPSqe+q+1ov1f9a+gBIltpfg+Ulsb7YlmY5aRymOvcnpUdxayzaDc3l+B5gs38mHORCNh592Pc9ug9wWsk2gyXl+B5i2beTEDkRDZ192Pc9ug75bqNxLf6FdR2bbLaK2fzZh/GQp+Rf6n8Bz0I35tHrzav8AT+vyAXWLiW/0a9js22W0UD+bMP4yFPyL/U/gOej/ABNNBZ+HJ7WJMM8RWOKMdFHU+wFP1yeGy8OTWsMeWe2dY4ox0ULyfYD/AOt1NR6xZ/ZvDWoyzP5tzLF+8lIxn0AHZR2H9aVLlvTb25tPPbX+vRA+p11FFFeYWFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFUb6+aJxbWgD3LDPP3Yx/eb+g7/maTdioxcnZCX1+YpBa2m17p13YPSNf7zf0Hes6SBbe4seS7tcMzyN1c+U/JoggW31XG4uzQFnkY8udw5NQ31xZzLHcX8myyjf92BndcvgjaoHJXkj3+lZvuzsjFJ8sdrav1T/pIuJB/a5w4xYKeT/z3I9P9n37/Tqk2syXUrWXh6JJ5E+V7huIYfxH3j7Co1s7/XFH28Np+nY+WzQ4kkH+2R90f7Irbt7eG1gWG2jWKJBhUQYAqkmzKcox07dOnz7v+vIoadosVnMbq5ka8vnHz3MvUeyjoo9hVLxaJymki0KLMdQUKzjIX91Jzjvj0roK5/xZcR2iaTNMSETUFzgZJ/dSYAHcn0qkrGEpOTuyjfx22i/2ZI7MR9tZpJGG55XMEoycdSTgAfQCmyQXE2taVe3+UczuIrcHiJfKfr6scDJ7dB6lZYbmbVtKvb/KM10yx24OREvkyHn1Y4GT26DuTHqcx1bV9PtbZnS2W4dZbhGxuby3yin6ZBI6duc4ZIalMdW1qwtrZnS2SWRZbhDjcdhyin9CR07HPRkOnef4l1GzjKw2Kw24kijGC4AbCey9c+vT1qzqsyaffaRb2UCvLukWG3UhRjyyM+yjuapWct1ZeItTt0P2vULiOBtxGEXhssfRRkDHU8dyTQBc+1m21u9tLCJXuZFiEceMJGoU/M2Oij9egpdNjh0q61aa7n3MGjaa4k6sdg/TnAA+lJp0cOk32rTXU+5sRPNcSdXO0/8A6gB9Kj0u3fUNevr29jaNEeNoLZ/4TsGHb/ax0Hb60ALpNu2oa1f3t4jxqk6NDbP/AAHy1w7f7WMcdue9M8JWLSaLYXd6wcxxYt4x92IdN3ux9ew4HfL4pbi81rVbOxYxp56eddL/AADy1G1f9rg/T64qt4debU/DtlaWxaK1jiCzzjguf7iH+bdug56AEmkyy6r4es7W1Zo7WO1RbidTguQozGh/m3boOek9veLZ+E7CGCPzbmWyQRQr3+QZY+ijuf6mlsL0WfhXTIYI/NuZbRPKhBxn5Rlieyjuf6kCk0a2hsfCkd3PJukkskaSaQ9FCcL7KP8A6/U0AJp1rHZeEzdXEm+aSyDSzOegCcKPRR6fj1NQCGTVfDfn3aFLOO0zDAf+WhCcO3t6D8TzjBaQS6p4chmvEKWkVoDDAf8AloQnDt7eg/E84xJd3ks3h5rLT8GRbLM8pGVhXZ092I6Dt1PbIAuqXskvh+eysAGkWyJnlP3YV2dPdiOg7dT2y/VRaaT4Luogdge1dFzy0jsh5Pck9SfqaLz7LpPgmWMfKHtWVR1aR2Q/mT1Jpl5ZSHw3f32obWumsJAiA5WBdh+Vff1Pf6UALqNnIdC1C+1Ha10bWQIgOVgXb91ff1Pf6VHrssmsaPefZ2Menxod0i8G4I7L/sep79Bx1Nbml1bQ742ztHYRQuTKpwbhgOi/7Gep79OnWx4nvFttFns7SMSTvD8sa8CNOm4+gH6ngUAHie8Fvo8lnaRh7h1XZGBhY1DD5m9F/n0FJqtvDpXh25mnlLuzxvPcOPmc71/TsAOlGq28Ol+G7yWeXfI+1p536udw/TsAOlRanBNfWBvr5WiRJIzb2x/g+dfnf/ax27fWgB/h21a7so7q6OY0mlaCAjG0+Y2Wb1bqB6fWnadNLeWQs7JjGqu4nuB/B85+Vf8Aa9+31qPSJZrvT/sdoWiVZpvPuAPuZlY7V/2iD17fWrGmTw6ZoNvFDGWcu6QwJ1c72/ySa9OrfmlfV30XbfX+vVkINMnh03w/axxR7nbcsMKdXOT/APrJo8O2og0mG7uZA87Rcu3AjT+6PQfz60eHbUQaWl3dOHnZW3OfuxruPyr6D+feodGjfVdKtfOUpYoowh6zkdz/ALPt3+nVVLfvEnpfV/fov69QXQNHik1TR7MTq0dlHCoEbcGcgdT/ALPt3+nV1rdyHw/ZWdiA1y1lGWYjKwrsHzH39B3+lOsLuT+xbGzsArXJtkLMRlYVx94/0Hf6Uumm20zwjbykbA9sjNjlncqOPUnsB9BTnu7r7Wi+/V/1r6AFoLbS/CMbn5Q9srMerSOVH5k9AKjtbWSbQ1vL7BkW0/cxA5WIbOvux7nt0HfK6baSS6JDd3uGdbQCCMfdiGzr7sR1PboO+YzNJqWgeTasUtY7XE0w/jITlF/qfwHPR68zs/tav9P639AC5mk1HQJYrRtltHakSzD+MhOUX+p/Ac9J9WuIrPwy9tEmXe0ZY4kHRQnJ9gP88mi+uY7PwwLeFN0slmRHEg6DZyT6Aev9TRNaC18MXs0z+ZcSWbeZKe/yHCj0A7D+ppK3utrTm0Xfz/r0QBe2gtfDeoSzv5tzLav5kpGM/KcAegHYf1qtrqy6joN5czhorZIyYYTwXP8AfYfyH4nno/U45NS0K7ubhWjtkt3aCFuC5CnDsP5Dt1PPR3ie8aTSL21tFDskeZ3P3Yx6e7H07dT2zVLm54d+bXy2/r8ED2OvoooryiwooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiqN9fNE4trQB7lhnn7sY/vN/Qd/zNJuxUYuTsgvr5onFtaAPcsM8/djH95v6Dv+ZrO06Lybi9Uu0jeapZ3OSx2KSaht7y307S2uL6b5nlfc7ctIwYgcdzgDgVVsLK88RSzzzmSz0uZw3lg4knwoGCey8Vne7TOxQ5Yzjslpfu7/1YiuLubWdWaHRo/tCIvlyyMcRdc8nuPYcn6dd/TtCis5hdXchu73GPOcYCD0Reij6VoW1rBZ26wWsSxRIMKqDAFS1SjbVmNSu5LkjokFFFFWc4Vh+JBmXRs8/8TFf/AEVJW5XP+LYPtKaTCXeMPqChmQ4OPKkyAe2emaAM3VJ/7V1LT7W3LrbLdsklzG2Mt5MmUU/TIJHTPHOcS6vKlhdaPb2MKvKJXENupC8CJxn2UZGTS6xItjNo9vYwo0q3B8m3UheBDIM+yjIyaZPapp99p93ezK87zO09w3AAEMnA9FHYfj1oAJLVNP1Kxu72ZXndpGnuG4AAjbgeijsPx61HpmoRHWtavrhTbwCK3IMgwSuHwcdRnsOvI78UXCyatrWlz3UeyyErmGF1+aQhCQ7A9PYfifQPtbWC78Z6lLIfM+zpAVTPyh8NyR6gdPTNADdMgfUfEF7eX0bRrH5TQW7fw/KcOw/veg7Z9aWOae917VLOxYxoHj8+5X+D5B8q/wC179vrSLNNfeINTtLB/LQeULi5X+D5T8q/7Xv2+tPt5rfR7u+gt4csZI0gt0+9I3lj/wDWSfqaAC1mt9HuNQgt4dzGdEgt0+9I3lJ/+sk/U0zQ74WnhXTIoI/Oupof3UIOM+rE9lHc/wBSBUui2xi1TVbu/MbXZkQPIOFRfLU7RnoB698ZNR+E7e3sfC9veTP8zwhpJpD91RnA9gP8T3oAf4etoLDwvBdzyZd7RGkmkP3VC8L7KB/j1NV9LtpdS8PWcl6my0htE8qA/wDLQhB87e3oPxPOMO0O2l1HQrCS9TZaw26eVAf+WjBR87e3oPxPOMFtdyz+HbSx04jzRYx+dMRlYAYx+bHsPxPuABvJZvDsVjpxHmrZL58xGVhHl9Pdj2Hbqe2ZXa10jwSFA2h7XCqBlpXZP1J60kT2ukeCYcLjfajaijLSuyZ6dyabb2T/APCOPfX5D3BsSEX+GBNn3R7+p7/SgAaykPhu5vtQIa5Ni4RR92Bdn3R7+p7/AEqLVZ5dV8O3ptnaOwhtHJlXg3DBDwv+xnqe/Tp1LyWXVvDc4gZo7CK0bdIODcME6D/Yz1Pfp062NevEtvDdxZ28fmTPZPiNOBHHsILH0A7ep4FADvEt4ltoNzZ20fmTPbNiNOBGmMFj6AdvU8CjVbaHTPDF/JNLvkkTdNPJwZG6D/AAfSjU7WLTfCupPLLvlkgYzTycGRiMD6egHaodWt5dQ0i5vdQQxxpGWt7Zv4P9t/8Aa9B2+tABqsE1/pkt7fo0aJg29s38HzD53/2vbt9aXxFcyX1jPZ2JIjR0W4uFONh3r8qn+9/L60eI7qS90+6s7A7UjwLi4H8HzD5F9W9fT61Lr72+leHRb20Y3F0WCBTgud4P/wBck0AGkXMen+H4Y40Lu00yQwqfmciVuPp6k9Kf4etRDp32u6YNcMX3uT8qKHPyr6Dv796b4atki0w3lwR57PKHYn5UAkbIXPQZ5qPR421bT4zKNtgruVT/AJ+DvJyf9n279+OvpVLfvEtr6v79P6/IhdA0aN9W0yAyqUsFyVQ9ZzuPJ/2fbv346yafdyf2TaWdiFa5MQLMRlYV/vH+g7/Siwu5P7Ngs7EK1yVJZiMrCu4/Mf6Dv9Kfo32bSvDUEjcBlDMcZaRz+pJ6AUVPtXXXRffq/wCtfQEGjm20vwvbSt8oaJWYjlpHIH4knoB9BUejWbyaLaXV4Q7JaqIIx92Ndg592I79ug75Xw/aPLpdld3pDMsCiCMfdiGOvuxHU9ug75h0+STUvD9rb27NHbR2qLNMOC5CDKL/AFP4Dno57zs+ur+/+vP0AIJZNS8PxQWzFLWO1UTTDq5Ccov9T+A56TzXSWnhiGCFN80lniOJfTZyT6Adz/U0LdLaeGLSCBPMnks1EcQ/3Blj6Adz/U0Wtsln4YkuJn3yyWeZJW9NnAHoB2H9TQ7btac2i7/1/wABAItqtp4YuZ5n8yeSzJklb02cKPQDsP6mobyOTUvD9xcXKtHbR2rNDCeC5CHDt/QfieehJDJqXh957pSlrHakwwn+MhOHb+g/E89JNWu3l0S5tbIBmW0YzyH7sa7OnuxHbt1PbLjzc673+7b+vwQDtdvHk0m9tbMBmS2YzyH7sa7enuxHbt1PbKeIFttN8J3UKnYGQquTlnY9/UnqSafqwttM8KXMSnYHgdVB5Z2Knn1J7k/U1BrNpIdBvry+w1y0JCqDlYV/ur7+p7/SlS5bw7c337f15AzsaKKK8wsKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiisjXddh0iDG5fObA55CAnG4jv9BycfWk2krsunCVSXLEm1TUmtI3jtVElyEL4PSNf7zf0Hf86w01MgfZtJj+2XrgNPKx+RGI6u39B6Y4qu1vfahYzuxksrQoztu/19wcdW/uj29OOlaml2Ed1ZxQwRiHT0Ubgox557j/d9T3+nXO7bO3khGFvPX+vyX3mfoegfbWF1flpQGbMjdH+Y8IP4U9T1OfSuwUBVCqAABgAdqFUKoVQAAMAAdKWrjHlRy1qzqyv0CiiiqMQooooAK5/xbJLEmktbxedL/aChEzgE+VJ1PYetdBWD4olSH+yZZnVI0vwzMxwAPKk5oAzrm1TT7rTru9mV53umae4bgACCXgeijsPx61HcrJquraVPcx7bL7QxhgkXlyI3IdgenTgfn6BbrzdU1LSprmPbZG7PkwSLy+IZCJGz05AwPz9A/Wbp7jWdOsbCTZOszGSbbuWHMT8f7xGSB+fuAJq909xrun2OnybZ1d/Nm25EIKH/AMex0H0J4qtaWU0PiLUrDTiYIWigM1wWy4Hz5xnqzEnntyfSrd/9m0WbSxGjsolkwqjc8rlD+ZJ7mqtlfTWevao16gku51gEVtCc5OHwoPsOrdOpoAswzW+j397Bbw7mPlJBbp96Rtp//WSfqado9qYdX1S7v2ja6JjDyDhUXYDtGegHr3xk0mkWrQ61qd3fuj3TLHvkHCxrtJ2rnoB+vWoLFP7c1jUHYkacJIyIyCDcHYME/wCx3A7/AE4IAWCf23qmoyM3/Et89GEZBBnPlryf9jgEDvn06t8MWsuoaFp73q7bWGMeVB/z0YH77e2eg/E9sWEu5RrOp2WnqDdPMhLMMrCvlJ8zf0Hf6ZNV9Bu5p/D1hYac2JRAvnT4yIAf5sew/E+4BJp13NceH7Gw01sSizj86fGRACg/Nj2H4n3ksZbXSfBNoQm3faptjQZaV2TOAO5J/wA4pdKmttJ8G2LBMeZboVjjGWlkZc4A7kn/ADim6JYsPD9teXjiWf7Eqx8YWFNg4A9T3Pf6UAJptk3/AAjqXt6wkuDY4jH8MKbOg9z3Pf6VC7y6v4bZIi0VhHaEO44NwwT7o9Ez1Pfp06pCZNY8NJEm6KwjtAJHGQ1wwTlR6L6nv0HHWxeXgt/DEdnaRiS4ex+WMcCNNmNzegH6ngUAGq3i23hlrO1jEk72JxGvAjTZgsfQDt6ngUt7aRad4O1BpJd80to5lnkwDIxQgfT0A7USW0Om+DbuSWTdLJZkzTvwZGKYH4dgKg1C3l1Dw3eXuoIY40s5Gt7Zv4P3Z+d/9r0Hb60AO1a3l1DQ7y91BDHGlu7W9s38Hyn53/2vQdvrTvEd3LeaZe2mnnCRIftFx2X/AGF9WPf0+tHiK7lu9Iv7XTyAkULfaZyMhePuL6se/oPepfEMltpXhiW2hjxuQpDCnVj3/wASaADxBJbaV4ba3t4xlsJDAnVzuBP+JJpuo2Zt9JuL3UJFe6cx75OixL5inaueij179T7Go2Zt9Fu7y/kWS7dBvfosa7gdq56KPXv1PtYt7VvFEizXKFdGRg0cTDBvCDkMw/uZ5A79TxigCnoUTappiGTH9niWVlUH/j4zIxyf9kenc9eOs9hdyfYI7OxCtclnLMRlYV3t8x/oO/0pmmXcn9nizsApuDNMWYjKwr5r/Mf6Dv8ASptHNvpfh9ZHJ5kcs2MtK28jt1J4FenVveV19rRffr/W/oQg0b7NpXhyN3OASSzYy0jFiPxJ6Uzw9aPLp9pd3uC6x/uYx92Ievux9fwHfJ4ftXmsbe8veXUN5MXaIZPPux9fwHfMelSSanpNvb27NHaqgE0ynBc/3EP82/Ac9FU/5eWfXV/f/Xn6AugaVJJqei2ltbs0dqkKrNMpwXOOUU/zP4DnpLbXYtvDNhFCnm3EtonlxDjPyDLH0A7n+tLYXYtfD2nwwR+bcSW6iKIHGeBkk9lHc/1pNHt4bLwzFdSvl3tEaSWQ9FCcD2A/zyaKlvebWnNou+/9fggDTreOz8Mi5mk3SPaK0krnoNnAHoB6f1NQRQyaloCzXSlLWO1zDCf4yE4dv6D8Tz0Wygk1HQYJbtdltHbKYoD/ABkJ99v6D8Tz0dNdSTaELOxxvW0HnSkZWIbOnux7Dt1PbNO/M7PW+/b+v+AgFvruSXQ5LSxwXW0zNKfuxDZ092Pp26ntl9/9m0zwjNGPlD2zKo6s7lT+Z7k/U0XBttL8IMg+UPbEKAMtI5T9Seppk1o//CP3d5fENcGycKo+7Cuw/KPf1Pf6VMeXTtzfe/6+4Bb20k/sK+vL7a1y1pIFUHKwrtPyr/U9/pVlIm8SSFRldJRvnccG6IP3V/2Aep79BSQRP4jIVSyaSnDuODdEfwj/AGPU9+g4ro0RY41SNQqKMKqjAA9KynUdNJP4vy/4P5euztcdRRRXEUFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFAGZrWo3FlFFFYQia7uGKRBzhVwMlj7D0rB/sNbVFvb6Zru+a4iJlbopMi52jt6ZrZ1f/kL6X/vS/8AoFNS3GsHDj/QFYEn/nuQc8f7II69/pWT1bO6m3TjFrRbv73p+GwsUR1Vu4sQeSP+W/sP9n37/TrsKoRQqgKoGAAOlCqFUKoAAGAAOlLWiRySnfRbBRRRTICiiigAooooAKwfFEUczaOkqB1/tFTgjjIjkI/Wt6uf8WpNImkpbSiKRtQUCQjO391JkgeuOlAFHWrlrnVdOsbGTZOtyS8wXcsWYZOP94gkge3PuuoC30ZtKEUbsq3D4VRueRzE/wCZJPU0aktvo66T5Ubsq3jHao3PIxhlH4kk9T6017Sb+2tLvdQINy0zqqKcrAvlOdo9TwMnv9KAA2k39u6be6gQblmkCxqcrAvln5R6n1bv9KbpNxa3XifWLyPaVSKFVmI4K/Pkg+mR174qO+f+29Y0+NARpwkkVpASDOdhyF/2OxPftxzTo9PW78VajE7bbSOK33QKMCTAbaD/ALI9O/HbigBtmn9ua5fuxI04eURGRj7RwcE/7Ht347cVMLuUa5qVlp6hrlmjJcjKQLsHzN7+i9/pk0G7lHiDULPT1DXLLFlyMpCu0/M3v6L3+nNLYfZ9Gn1RpZHYCSMs7fM8rlB+ZJ6AfhQAunG30Z9VMsjsFuEy7nc8jGNPzJJ6D8Kj8PT2+meDbB9nMiDbHGMtK5J4A7k/54FLo9s9zrWoX1/HsmWZfKhLbhFmJef97GATzjkD3b4QscaHZXly/mzeTti/uxJ6D3Pc/wBBQA/w5YkaHZ3l44lm+yKsXHyxJtHAHqe57/Sq1h5mseGraEborCO0RZG5DXDBBlR6L6nv06dV0gSax4dsoAGisI7ZFlbo1wwUAqPRfU9+nTrNDfND4XsbWzQS3klghRD91F2D529AP16CgAe9MHhe2tLOMS3Ulgu2Posa7Pvt6Afr0FOjgg0zwbNNLJl5LPdNM55c7MD8OwFFnDbaX4NNxK+C9mHmmc8sdnA+nYCoY7WW+8O/bNQXbHHZk29uf4fk++3q3oO31oASa2lv/Dkt7qCFI47Njb2zfwfuz87f7XoO31p+t3kt1oN9baeRshtH+0zkZC4Qnyx6se/oPek1G7lvPD9xa6eQEitG+0z4yF+T7i+rHv6fWpdYktNK8GzW0a7A9o8cUSDJJ2HJ/DqTQA7X5LXSvClxbRJt3wNHFEgyWOOT+HUmm6lZNBod/eXzrJdyQne4+7Gv9xfQD9ep9jULJofD+pXd66y3klq4Zx92NcfcX0H8+p9rFvbP4nkWa4Vk0ZDujiYYN4R0Zh/zz9B/F1PFABb2zeJ5FmuUKaMjBo4mGDeEdGYf88/QfxdTxiunAwMDgUAYGBwKKAOW0mW3sdFknkAQNczFto+Z281gBjueAKZ4ftXmtIry85cNJ5MXaEbzn6t7/gO+V0Gz3wm6nfzNk84hTHEf71sn3J559OPXMemNLqNittAWjtVdxNMOC/zn5UP82/Ac9PSqb1LPrq/v0/r02IXQTS3k1LTYre3Zo7VdwmmHBf5j8iH+bfgOek2n3Qs9BsYLePzLiSPEUIOM+pJ7KO5/rRp90tnotpBbRiSdwwihBx/EeT6KO5/rS+H4IrHQorq4kBkaINLM56AdvYClUtaV1pzaLvv/AF+CBBoMENl4fhup5Mu9urSTSHooHA9gP89ag0u3lvtDs5LxdltDbJ5UJP3yFHzt/Qfieei6JbyX+k2Ul2uy2iiTyoT/ABkD77f0H49eiWd1LPoNpZ2BxILOPzpsZEI2Dj3Y9h26n3qd+adnrfV9t/6/BAKl1JNoMNnYEeYtmvnTEZEQ2Dj3Y9h26ntmTfbaX4PQAY3W2FVRlpHKfqT1ogkt9M8IQYXG+2G1EGWkcrngdyabaWjDQftl4wec2eEH8MK7Og9/U9/pQ+X5c33/ANfh6gIto/8Awj815ekNcGyYKP4YV2fdHv6nv9KntIn8Qxog3R6UgAdxwbojqo9E9T36Dim2EMniG2hj+aPSo0VZG6G6IHKj0T1PfoOK6dEWKNUjVURQAqqMAAdgKzqVHT3+K/3f8H8vXZpXBEWONUjUIigBVUYAHoKdRRXAUFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRUNxcCAABHlkb7saDJP9APc8UDSbdkZviDTbi+S3ltAkj27FjC7FRKpGCuR0qfTNXtr/ADAqtbXMQxJayja6fh3HuKf5uqMcraWqr6PcNn9Ex/Oqt3Zw6qyLf20lndp/qbiJ+VP+y4/kR+FR1ujps3BRnsuzWnyNiisJdUvNHcQ68PMtycJfxr8v/bRf4T79K20dZEV42DqwyGU5BFUncwlBxHUUUUyAooooAKKKKACue8W3AtY9JmZHfbqC4RBlmJikAA+proaxPEn+t0f/ALCA/wDRUlAGXLazf2rpV7qDA3L3TKEU5WFPIlO0ep4GT3x6AVFqEn9t6pp0Sg/2d57qZQxBnIjfIX/Y4IJ7544o1GUa1qGnRKp/s77UymUMQZmEMmQv+xgEE988e8+tXPkalpNpYLG115jGOI8Ki+W43HHQDPTvjAoANYujBq2lWlgkbXRMhjiPCouwjccdAPTvjAqlZrdWfiLU7S0Zp7ueOBnuZR8qcNlj+YAUfyBNW7iG30i8sZ7mbc7SSPPcSdXPln/9QA+lQ6fqPlazrF5fRtbqUt9kZ5cghwowP4j6D1xQBPZC20W71NpHYj90WdvmeVyD+ZJ6AfQU3SLV7nXr++1CPbMrp5UJbIhBQc+m7HU9uQKNKtpLnxDfXuoR7ZlEflQ7siEFT+BbHU9uQKZA8mq61qlvaybLQSIJ50b5nwgBRSOnue3QeoAC2aXVNX1WC1k2WfnqJp0bl8RqCikdOnJ7dvUQeHEl1bw7ZW5DRWEcQWVujXB7qPRfU9+nTrbguY9Ovr+1soVeZpkWC3TgYEScn0Udz/XAqLRL5ovDOmWtmglvJYAVQ/dRf77eg/n0FAEljfND4a021s0WW8ks4yiH7qLtHzt6D+fQUulR2uleDY7mV9u+zSSaVzkn5BgfQdAKXQo7XS/CMFzK+3fbJJLK5yT8owPoOgFQ6RaS3eg2dzqAAjhtE+zwA5Awn+sb1Y9vT60AJZWkt74fhutQXbHFaA29vnIX5Pvt6t6Dt9aLi6lvfDz2mnttSOzxcXHZfk+4vq3r6fWgXMt94dSz09yiR2ai5uR/B8n3F/2vft9amnmt9L8GRwxR/NJZ7IYUHLnZk/4k/jQAahLa6V4MNvEm3fZskMKDJY7OT/Umm3Vi0XhfUbu9dZbySwkDMPuxrsPyL6D+fU+wbIw+F7u7vXWW7ksGDP0WNdn3F9B/PqasWNrJ4lWKW5Ro9HjAMcTDBuyP4mH/ADz9B/F1PFADrW2fxM6zXCsmjoQY4mGDeEdGYf8APP0H8XU8V04GBgcCgDAwOBRQAUUUUAcdpfn31k1pHvht0nm86UcF8ysdi/geW/Ac9J9PuVstHt7e1iDzu0ghhXjo55Poo7n+tJp179m0pIoU825lnn8uIHGf3rZJPZR3P9adoEEVlpDXVy6+axczTMcDAduBnoB6V6Vb7d1pzaLu9f6/BEINAgistFF1cSDzGUtNMx6AE8ewHpUOhwSajplnLdLttY1BihP8ZH8bf0H4+mDRIH1HT7ea6G21jJMMJ/jO4/O39B+Pph2nXMs+l21lYNiQRjzpsZEI/qx7Dt1Pu6l7z11vq+2/9fggXQTTrmWfR7OysGxILdPOmxkQjaPzY9h26n3fYy2+m+EbMhMeZbJtjQfNI7LngdyadpctvpnheyIT78S7Y0HzSORnAHcmmaJZldDtru6cSS/ZFEfGFiTYOB7nue/0pT5fe7c337/15eoITTLNhoUd3dsJJjZgJ/diTZ0Hv6nv9KdpkMniCxt4zui0uONVkbo10wHKj0T1PfoOKbpEEuv6baxEPFpcUKLI3KtcsAAVHogPBPfoOK6tEWONUjUIigBVUYAHoBU1qjpyafxX+7/g/l67CVwRFijVI1VEUAKqjAAHYCnUUV55YUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABUFzeW1kqvdzJCrNtDOcDPpmp6guZYIthugAmeHYfKp6cnt1NJ7FwScrMpm+MZlbzWuRKwFuluqk4xzg9D756Y9xV62eSW2R50MbnqpAyOeM4J7VUfRtNnYTRwLG55EtuxjP1yuKmgliiK2scz3Ei/eJbcyj/AGj/AJNSrp6m1TklH3Fr10LLKroVdQysMEEZBFYb6Vd6Q7T6AwaEnL2EjfIf9w/wn26Vu0VTVzGM3EoabrFtqW5I90VxHxLbyja6H3Hp71fqhqWj22pbZG3Q3Mf+quYjtdPx7j2NUU1a60mRYPECjyicJfxriNv98fwH9KV7blcinrD7v63N2ikVg6hkIZSMgg5BpaoyCiiigArn/Ftsl2mkwTbvLfUFDBTjI8uTj6HpXQVz3i1p1j0n7IivN/aChA5woPlycn2HWgCprdx5N9pNpYCI3X2gtHEThVXyZFycdAM9O+DimXUNvpFxp89zNudrh3nuJOrnyX//AFAD6UXkNvpEmnXFzNuc3jPPcSdXPkS/p2AH0ps0c97rOlXd6pji89/JtWH3R5TkO/8AtcDA7fWgBJIp77XdLu75DHGJJDBbMPu/ITvf/a9B2+tOsIra68Z6pPxK9ukKqc5CNhs8dM9s9uR603Vbh7/XbGysZGjVJJFmuU/gOw/Iv+1jv2+tQ22nv/wkWo2NmRbWYhg80oSHIw/yg9s85br+JyACWJ5NV17UoLSXZaAxrPMh+Z8Agop7dwT26DnkSRXMenajfWtlCrzM0awW6cDAjHJ9FHc/1wKEuY9O1W8tbKFXmZYlgt04GAp5Poo7n+uBTtJhWw1DVrm+nV5sxtNOwCgDYDgeijtQAaRCtjeaxc30yPN5qGa4YBRgRKceyjJxUfhZLXTPCNvdSNsDxh5ZXOSfQfh0ApumQnVtX1C5uA62izo0Vu643t5aYdh+RAPTvzjDPCllJdaJp9zf48uGMfZ4AcgH++3q3p6D3oAfoFnLd6LYXOoACOG2T7PADkDC/wCsb1Y9vQe9MtLiW/8ADdtZae5REskFxcr/AAfux8i/7Xv2+tLpdxLf+H7Ky09zGiWsYuLlf4PkHyL/ALXv2+tTW11Dpvg2xSOPfLLaKIoIxgyMUyf8Se3WgAFxBpngy2jijzJJZhYYU6uxTJ/xJ/Gkt7LyPDE15eSCS5exO6Q8LGmzO1fQD9TyaNNsxb+Gftl1IJLh7EZkbgRpsyFX0A7+p5NO0izl8RWdq92jRaTFGmyFhg3bAD5mH9wHoP4up4xQA/TLSTxHBbyXSNHpEarshYYN2wH3mH/PP0H8XU8V1YGBgcCgDAwOBRQAUUUUAFFFFAHLaFHb2umz3bkIWmmMkjnoBK/HsPb3NQ6JA+o2MU1yMWqSO0MJH3zvPzN9Ow/H0wujWkl3b7rsj7NFczGKIfxt5rHc30PQe2fTC6fcyz2S2Vg2JA7+dNjIhG8/mx7Dt1Pv6dS/NOz1vv2Wv9fgiF0DT7mWfT4rKwbEgB86bGRCMn82PYdup95NJlt9L8NWmFJLLhUUZaVz2HqTRpUkGl+HbfhiSSFReXlcsePcn/PFJ4dtCNNt7y7YPMYsIP4Yl9B7+p7/AEqanLafbm+/f+vL1BCeHrMrpNrd3biSb7OoTjCxJjoPf1Pf6VHolvLr2kWULB4dMigRJTyrXLBQCo9EB4J79Bxmk0S3l17SbSFg8OmRRKsp+61ywHKj0Qdz36DvXXIixxqkahEUYVVGAB6CitUdOUr/ABX+7f8AH8vXYSuCIscapGoVFGFVRgAelOoorziwooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAoIyMHkUU132IWIY47KMmgCo2jaa7bjZQZPXCAA/hVqGGK3jEcEaRIOiooAH4CsxdegdZBtuVZXKgi0kOMevHWtC0lee1SSVCjNnhl2nGeDjtxzioi430OipGso/vL/MmoooqznCmyRpLGySqrowwysMginUUAYTaZeaMxl0I+dbZy9hK3H/bNj90+x4q/pur2upqwhLJNHxLBINskZ9x/Wr1Z+paNb6gyzBmt7uP/V3MRw6+3uPY1NmtjXmU/j+//M0KKw4tYudMlW28QoqBjtjvox+6f/e/uH9K2wQygqcg8gjvTTuTKDjuLWB4rnitV0qadwkaX4LMe37qSt+sPxMqvJowdQw/tFTgjPIjkIpkGXOlxeappN1eqYojdkQ2rD7o8mQh3/2sgYHb65pNWuGv9ZsLKyd4wk7ia5THyHynyi/7WM89uKXV7g3+qafZWcjxhLoiW5THyN5MmUH+1jPPbinas8ekyaRFZWxldZ3ENtEfmc+W479skEsenU0AJqkkWkXGkxWVs0rh5FhtovvOdh7ntk5LHp1NVLKa703xBqcN0y3ep3McDLFHwgOH4HoijGSf5nFdJo+jvaSPfai6z6jMuHkA+WJf+eaeij8yeTWeHig8Wa1PMyoqQW5Z242jD55oAg0uFbDU9VuL6dXl2xNNO3ygDaTgeiiodOiOr61f3NyrpapJG0UDjG87Bh2H6gH6kZxg0+I6vr17cXKOlqhiaKBxjzDg4dh+oB+pGejhLeXev6lYaV8js8fnXRGVt12Dp6uew/E0ACS3l3ruqafpXySNMhmuiMrAvlIOPVzg4H4mq/h+4mvvD1lY2DmNEiAuLkfwf7C/7Xv2+tdfp+n2+l2a21ohVFOSWOWdj1Zj3J9a5fQruLT/AAhpoSPfLLHiKCPgyMck/T1J7UASaddRad4R01Y4t8stsgigjABkYrk/4k9utJolmtv4ahvLmTzJnsVzI/AjTYCFHoB39TzR4ds0tvDtte3Mu+ZrNMyPwI49uQo9AO57nn6M0Cxm1/SbD7dG0OlQQRhYW4N2yqPmb/YBHA/i6njFADtEspvEGnWTXsbRaVDEgWFhg3bAD5mH9wHoP4up4xXYAYGBwKAMDA4FFABRRRQAUUUUAFFFFAHH6Xcyz2RsrA7ZBPN502MiEGV/zY9h+J959Klg0vw/FwzEyOqIOXlbe3HuT/nijTLqKy0XzHBLPczBUQfNI3mvgAdzSeHrU/YkvLvmcmQKpOVhXecgf1P+FelVtad1pzfe9f68vUhdA8O2h/s+K8vDunKsFGflhXceB/U/0qHRbebXdMt4mDRaXGuJG6NdHuo9E9T36DvS6NbTa7p8UTBotLjyJGBw10cn5R6J6nv09a61EWONUjUIijCqowAPQVNao6c5X+Jv7t/x1+XrsJXBEWONUjUIijCqowAPQU6iivPLCiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAzpF1fzpfs5so4tx2B42LEepw1WNPadrJTdurzbmDlBhc7jwPYdKyL22tGvpimmXV8+7MsiSFVU46D5hk+wrZsvI+xx/ZU2RY+VcYK+oI9c9azjudlZJU1p26L/Nv77E9FFFaHGFFFFABRRRQAyWKOeJo5kWSNhhlYZBH0rEOn32hsZNFJubPOWsZG5X/rmx6fQ1vUUmrlxm46dClp2q2uqRs1s5DocSQuNrxn0Ydqy/FsLzx6TFHM0JfUFHmJ94Dy5M49Djv2q/qOiw30i3MTta3qD5LmLhh7H+8PY1zPiXWL/TYrD+1rJpJLe7EizQf6uYeW4/4CcsOPqaLvqNxi1eP3FrV2j0r+x4rG2MrrdMIbaI/M/wC5kHftlhlj0zk1saPo72kj32ous+ozLh5APliX/nmnoo/Mnk1g6Pr+g2kj32o6kJ9RmXDyC3k2xL/zzT5eFH5k8mtb/hNvD/8Az/n/AL8Sf/E0zM3q5R7FLzxtqTTszRRR27eT/CzYbBPrjsOmeewq7/wm3h//AJ/z/wB+JP8A4muXvPFFq/iS++wXfkxXiQqbxoXIiCht2FxktyMcYoA1fMvLvxFqFjpXyO3lia6Iytuu3t6uc8D8TXT6fp9vpdmttaIVRTkljlnY9WY9yfWsHT/E3hbS7Nba0vGVFOSWhkLOx6sx28k+tWv+E28P/wDP+f8AvxJ/8TQBvVxvhG0itvDtte3Em+RoOZH4EcY52j0Hc+p5rU/4Tbw//wA/5/78Sf8AxNchoeq6fd6daWmtXBt7G1RQbfy3Y3LjnLYBGwf3e568cVUYyl8KuBteHrCfXtH0/wC3RtDpcEEYWBuDdsAPmb/YB6D+LqeMV2YGBgcCsIeMtAAwL3A/64Sf/E0v/CZ6D/z/AB/78Sf/ABNX7Gp/K/uFdG5RWH/wmeg/8/x/78Sf/E0f8JnoP/P8f+/En/xNHsan8r+4Lo3KKw/+Ez0H/n+P/fiT/wCJo/4TPQf+f4/9+JP/AImj2NT+V/cF0blFYf8Awmeg/wDP8f8AvxJ/8TR/wmeg/wDP8f8AvxJ/8TR7Gp/K/uC6NyisP/hM9B/5/j/34k/+Jo/4TPQf+f4/9+JP/iaPY1P5X9wXRR0C0QW73crF386dU3dIl81sgfU8k/4VFo9tNrlmsTBotLR33sDg3R3k7R6J6nv09abodrPrdiIpFeHSxNKztyrXWZGYKO4TBGT1PT1rr0RY41SNQiKMKqjAA9BXZXqeznLrK/3b/j+XrtKV0CIscapGoRFGFVRgAegp1FFecWFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAZhj1dJ5vs7WawtISgkVicfgfXJosY9SSZBcS2skGX3GEEHdk9cn1zVK5OnSX1x9str2eRXxvjSUr9Bt4GK1NK8oaan2ZHSLc+1ZM7h8x655/Oso6s76l40723t08u9y5RRRWpwBRRRQAUUUUAFFFFABRSEZBHTPpWHD4fuzCpn17VVk/iCyx4/D5KANzcpYrkbgMkZ5A/yKWszTNLl0+9uJJL65vFljjVWuWUspUvkDAHHzCtOgCOeeK2gea5lSKJBlndsBR7moLXVbC9tGubO6inhU4Z423BT746UzVLX7TBEWtxcpFKJGgOP3mAQOvHBIPPpUEFhvWLULe0XT75kHmRcYcf3H28H2PUfmCAaME8VzAk1tKksTjKujBlYexFSVl+GrOfT/DdnaXabJokKuuc4OT3rUoAgmvbW3mihuLmGKWY4jR5AGc+wPWoF1rTW1M6ct9Abwf8ALHeN30+vtVa80yO61YfatLgvIZ12vcSFcwgA4UAjPJ9PU1DLpiQLa6fBpom+UOb7CjZIGB3nJ3bjy2fX8aAN2iiigAooooAKKKKACiiigApGYIpZyFA6knGKWuce70Zrh21lJHdJGQS3kDeTlSQdnVQMj6+ta06bn3+SuJuxvxTxTZ8mVJMddjA4qSsdbXQ7p4zb29s27/Vz2wAwfQOnIPFakEbxRBHlMuOjMOce/qfelOKjtf5gSUUUVmMKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAqK488xj7KYw+4Z8wHG3PPTvipaKBp2dzPddWM0nkNZJFu+QPGxYj1OGqzZrOtqBd7DLuYsUGAfmOMfhU9FJKzLlUco2sgooopmYUUUUAFFFFABRRRQAUyWaKBd08iRrnGXYAfrT6wteEbahaCXRW1b93J8mxGWMZXn5yBnoP8A9XIBrQXltcyPHbzxysgDMEbOAc46fQ1PWJoSxLfXYh0Q6T+7iyCqL5nL/wBwkcfnzW3QAVDLeW0D7ZriKNvR3ANTVysiQNfXZk8KPqH75v8ASWihJfn/AGmzgdB7CgDpoJ4rmESwOJIySAw6HBx/SpKo6Nt/smLZa/ZF3PiD/nn8x+Xjj8uKvUAFVW1OxUqDdwlmYIAJASSTgDA9zTdXx/Y15uR5F8lsogyWGOmMjOfrXPoluJLY/wDCHG0fz4v3wjhAjO9ecqc8fSgDrKKKKACiiigAooooAKKKKACq99A89qVh2eYrB1Eg+UkHOD7f/rqxWe9newN/xLrqMR9obiMsq+ykEED65q4b3vYRSs4Gl1aKd9BWxnjJ33AaMgggjAI5PbsP6HdrNWLWJTia5tIF7mGJmb8CxwPyNXoIVgiCIWbuWY5LH1JrStLmad/uv+oIkooorAYUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABWF4muIreENeXU9rbiNyrxOyAyfwhmXkew6HnPati584Wk32QKZ9jeUH6bscZ9s1ivH4i+0JbnUNNbejM2bNsYBA/56c9aALumahBqNzPNYStNa4X95zsL852k+2M44/HNaVZWmJqcWoTx6ndW8y+UhiWCIoF5bJwSeen5Vq0AFczc6pYWGtMdU1CaGZmdDDJIyp5ZHylFHBzxyMnOR7Vs6iuouIl0uWCIlj5jzxlwBjsAR3xWZFF4iudz/AG/TcJIyqTZtng4J/wBZxyDQBr6e7yWEbyhwWyVEgw23J2599uM5qzVLSTcHTY/tsqy3AZxI6jC7g5BA9h0H0q7QBXvjKthObfd5gQ4KDLD6DufQetc/p+rac7mz0y/nu7j7SpjVpHd9uF3ls9F+9wenbnFX9ROuRvdT2t1Zw20YzGkluzsw2jOSHAHOe1QSweIIfLkmv7DZ50fm+VasrFdwyAS5+nSgDoKKKKACiiigAooooAKKKKACmvIsagucAkKPqTgU6sjXrW0e2W4vby5tEjkUloZnAPIwNqn174zmtKcVKSixM02uI1nWJtwZumUOD7ZxjNSVzWnXOinVLdLa+vby4Yny1nlldU+U5bDcdOM+9dLVVafs2lr81YE7hRRRWIwooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKAI7mdLW1luJc7IkLtgZOAMmud1HVrPUFjjlsdbglGTFJBbSJIvrgjt0yOnSumIyMHkVQ1BL1WjutLSGaRFKtDK5UOpweGAODx6Y5oAo+HZoDNcwQwaiHVUaSfUAwkkJzgfN2GO3HP1rdrO077ZJcSzaksEM5RQLeFy+xcnktgZJOe2OPrWjQBT1HU49NWIyQ3ExlYqqW8RkbpnoPpXOXd1ZXt280dv4itmziZbSKRFc4H3gO+McjB6V1xxwTj2rJn/tmzuHTTrW1ureRyytLMY2iJOTkbTuGSTxz296ALWjzRXGkW8lvbvbQlSI43BDBQSBnPqBn8au1XsEMdmqtKsr7mLugwpYsS2PbOasUAZF/rtrBJc2s9le3CRjbM0dsXjwVBIJ6dDzWLbzWcF5A4tvEEwMyLFFdrJ5MZLAA8+mcjOeldXcw/aLWWFW2NIpAbGcHscd6yw+szyLDqMFlbW6yJvnSYuZPmGAqkDGTgcnjPGaANqiiigAooooAKKKKACiiigAqrqBuBaZsoIp5t67VlOFHzDkkZ6dfwq1Wfrk0MOlO11K0UBdFlZCQdpYAgY559ua0pq80hPYithqtxdRterYJFExJNvIzsTgjHIGOtatcpBfeGdOvIZdMkW3JJWTy0cKy4P3hjnnGD1/Wuks7uK+tUubckxPnaSMZAOM/pW2IpyjaXK0vS3+YkyeiiiuUoKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAIyMHpXOappuh2sdlBPpct2sKOsFvBE0u1SQST6c45J7mt65877JN9k2+fsbyt/TdjjPtmsdtN8QNMkv9sWauoI+WxPIPY/vPYUAN8Opp8d9eLpulT6d+7iLiaLy9/L4wPbnn3rfrO0231OG6nbU7yK6VkQRmKHywpBbdxk+o5rRoAgurO3vY0S6iEixyLIoJPDKcg1zV7a6B9vuDJoN1fytIxknjtWdS2eQG746cVv6imoyLEumTwwHcfMeaIyYGOwBHfHes+PTfEETPs1m0AZt237AcAnrj95360AXtD8kaPCLWB7eEFwkTjDINx4I7fTtV+q2nx3MNiiXsglnBbe6jAb5icgduO3arNAGdqOn6e8N3dXkX+stjDM65LGPk4AH1PTmufSDQ47m2a38NXlvJ9oi2TNamMIfMXBJPQVtX1prN3PcJBe2tvathUR7YyMRtGTkOMc5qIWPiBTD5ur20yLLGZFW02M6hgW53nHGe1AG5RRRQAUUUUAFFFFABRRRQAVm6tqkGnrGJ4pX3SIBtgZx1HoOvcfStKq2oC6NpiwWFp96488EqBkZOAR0HNaU+XnXNt9wmU7fWBeX8cdrHcFGzv820eMKMHnc2B1xxWoqKgwgCjJOAO5OTWQh1tL61+2z2P2cuQ4gjcM3ynA5J4zz+ArYq6yimuXb+vJAgooorAYUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFR3ErQ28kiRPMyKSI0xub2Ge9SUUAUIdb0ydNyX0CkfeR3Csp9Cp5B9jUn9q6f/z/AFt/3+X/ABqaS2gmbM0Mch9WQGmfYLP/AJ9IP+/YoAZ/aun/APP9bf8Af5f8aP7V0/8A5/rb/v8AL/jT/sFn/wA+kH/fsUfYLP8A59IP+/YoAaup2LsFW9t2J6ASrz+tWWdUUs7BVHJJOAKrtp1k6lXs7dlPUGJSD+lUovDlklwWkDTWy4MNpL80UJ7lVP4denOMZoAuf2pp/wDz/W3/AH+X/Gk/tXT/APn+tv8Av8v+NP8AsFn/AM+kH/fsUfYLP/n0g/79igBn9q6f/wA/1t/3+X/Gj+1dP/5/rb/v8v8AjT/sFn/z6Qf9+xR9gs/+fSD/AL9igCtNrlijLHbzLd3DnCQW7B3PvweB7nArRpkcMUKkQxpGD1CqBT6ACiiigAooooAKKKKACqOrw28+msl2sjx7lOyInc5DDCjHqeKvVDdQvPBiGQRyqQ0blcgEeo7jt+NXB2kmBiaRa2ttqCmTRZ7CZsiGSWYSg8HIyGO04z+Ga6Gs22g1SW5R9UktBHExZEtlbLHBGSW6deg/OtKtK8uaV27/ADb/ADEgooorAYUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFUNcu2sNCvLqN9kkMRZCADluwwfU8fjV+sfxFCb2Oy0/YXjurtPOGOPLTLnP12gfjQBfsbgTWyK1xFNOiKJvLYHDY54HTnNWaxbXcfFl45jZVSKO2jAGBtALl/plgo9wfQ1p3NvLPt8m7lttuc+WqHd/30DQAW9yLiW4QKR5Evlk56/KrZ/8AHqnPTjrWHp9jdG61HGqXK4ucEiOL5v3ac/crbUFUAZixA5Y9/egDN07ULm51bULS5jiT7L5e3y2LZ3KSeTj+QrRkVmjZUfYxHDAZxWFpU8beKNXYbts3k+WxQgPhDnB74rdd1jjZ3OFUZJoAy9B1UX2j2Ul3cRfa549xQEAseei1agupJdUuYfNtXhiVcLG+ZVY5zvHYdMfjXKWgWHwloKeW6SR30TyL5Z3LhjuJGOOD+RrQvFnuNY1tNPLLPNpyRwvggFwZOAemRkfnQBf1XVDDPYiyu4iXvEhmjBDEqTg+4xWjc3EUY8prqOCVx8hZhn6gHrXN6hPbDTNCSKNl+z3cBdPLOYAAQdwx8v41esT9k1HV/wC0gS9xOHiYrkSRbFCqvrg7vl9896ALOiakbrw3ZX+oSxo8sSs7khVyaum+tFjSRrqEJI21GMgwzdMA9zXI6XHMdJ8Mz7pEtrVHS4wmfKcrhWII6DkZ7bqs63bwroV95LzXH2q8gkPyZViHj3FAo6bVyT9aAOniuYJ2ZYJo5Cn3gjg7friq2tTTW2h3lxayeXLDA8ikqDyqk4x+FUEkQ+NgY8+W2nhAwU7Sd+QM9M4OcVc8QH/inNRXBLPbSIoAySSpAAFAE8V7EljbzXc8cRkjDEuwUHgE9aklvbWEIZrmGMSfc3yAbvp61gxzBtQsPl+zqthj7UyEliSAYlB4z8ueQT6DrVaws57vwUtokUkd7YkywGRcbZUkYqv5AD6NQB1H2q3854fPj81F3Mm8blHqR2FMGoWZheYXcBijxvfzBtXPTJ7Vg6hDcap4ba+ghdZLp4ZZIguXMCsDs9+MnHuRU9wlu/2q9S5luGNk8LEINnJyAcDJbOcDtk56igDYW+tHDFLqFgihmxIDtB6E+xp8VxDPu8iaOXb97YwOPriubhtbX/hEtPKpJBP5dvG08Mf72JlIbJBByA2SQfU1o6L9qFzdi6nivFATZeRx7PM65UgHBK8cj19qALep3b2lqvkYM80iwxZGQGY4yfYDJ/CrGUtrcedLhUABkkYDPuTWdrI23ekSk4SO+G7/AIFG6D/x5gPxpmthria3t1uJLFoz56XeFKBh8uwqfvZDE49s0AWtQuWTTWvbJxL5AMm1WyJFH3l+uM498e9XI5FmiSSM7kdQyn1BrJsWktfDM0+ooscm2WWYhSu8ZJ3YPIyMHHbOO1W9Fie30HT4Zv8AWR20avn1CgGgC7RRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFADVjRGYoqqXO5iBjccYyfwA/KnUUUAFFFFABRRRQAUUUUAFFFFABRRRQAVVurEXUyOZ5owqMjIj4Vw2M5Hrx16jJq1RQAgAAAAwB0ApaKKACiiigCK6to7u2eCbO1u4OCCDkEHsQQCPpT4w6xqJGDuBywGMn6U6igCvd2ovI1ikbEJYGRMffA5x9M9fUcVYoooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAoqtqJcaZcmKRonWJmV0xkEDPcEVU0bVIrnStPM84aeeBCWIwHfaCwB6E9eB70AalFZ/nPa64tszFobuNpI8/wOpG4fQhgcf7J9a0KACiua1+bVNM8N6jqCahMk8bM0SeXGVVd+FH3cn5cdTWr9lvEvExfzyQPGyuGWMFG42sCFHuO45oA0KKxPD+rpcaJYG9ud9zMNpdhjc/PGQMbsDp1rSn1G0tpGSaYKUAZ+CRGD0LEcKPrigCzRRVO51WxtJ2hubhY5FjMpUg/dHU+/WgC5RVSTU7OMKWnUho/NygLYT+8cdF9zxWbq1+ZNVstPgu3gS4hlkMsS5OQFCYOOnz549B2oA3aKppqFtGVgluQ8isImcrgF/QkcBj6e9NbWdPWZoWukWRJBEynIIY9B+OR+YoAvUVVbUrRZjE067g4jJwdoY9FLdAeRx15FMbV7FZJ4zON9v/rVCklPcjHT36UAXaKgkvraKGOVplKTY8or83mZGRtA68c8Uz+0bTyPO89du/y8YO7f/d29d3tjNAFqiooLmK6QtA4YKxVhjBUjqCDyD7Gm3l7b2Fq1xeSCKFfvOQcCgCeiq8d9bS3HkRygybSwXBG5QcEj1GT1FNGpWjTCITKWLmMHB2lx/Du6Z9s5oAtUVmWWt295eXkK7gLaTZkxsM/IGJ6e9SprWnSRxSJdoyTOY42GcOw7D1NAF6iqv9pWeZh9oTMDBJB/dY9B9Txgd8il/tC28qWQybRCMyBlIZB6lSM0AWaKx9Q1a3uNHvW02+AnjtGuEaPGduCQwyMEZGKs2mowmKzgnnX7VPAsgQ9W45P50AX6KK5Y+INVt4b68uYrJ7WxujBKsZZZGHy8rnIJ+Ycd6AOpoqpfapZaaqG/uo4N5woduW9cDvVfVNfsdM0hr9545EaNnhCuMTEDIAPvQBp0VzcOu30vhrUNSjfT5ntozInklyuAu5lbPftmpbXWr27utMVEt0S4SQXCsGLRyRnDgc+vTNAG/RVE6zpo1D7Cb2AXO7b5e8Z3en19utVLrxVpdnrAsLi5jRtjM7luEII+U+5zn8KANmimxyJLEskbBkcBlYdwehqG5vra0kiS5mWNpm2xg/xH0HvQBYoql/a9h5M0puUCQNslY5wh9D71Gl6p1adTeq0aW6v9m8o7l5Pz57g8DHtQBo0VmaXqket6c0kDyRFmcAhCCAHKg8jGcAVX8PawlzoenNe3Ie5uIwCzDAd8ZxnpnHOOtAG3RRVaTULWG48mSdVlxnaep6cD1PI468j1oAs0VFBcxXUZeBw4DFT2KkdQR2PtUtABRRWeZnutca2UlYbWNZJMH77sTtH0AUnHuPSgDQorDl1LyvFksUl0wtoLMSSRbcgMzHB4GeApz25FbaOsiK8bBlYZVgcgj1oAWimyypDG0krhEUZLMcAVWlYX1rKlrM0cy/dYqQUbGQSDzj2PUUAW6Kq6bef2hpsF1t2NImWT+63Qj8DkVaoAKKKKACiiigAooooAKKKKACiiigCC+RpNPuEQFmaJgAO5wawHtLm48M6XpccEsV1D9n3uyECHyypZt3Q/dwMHv9a6aigDLvR5viPS0XrEs0zew2hf5t+hrUqGK2SO4lnyWllwCx7KOij2GT+ZqagDC8Yxy3Xhe8s7aGaaeZAESOMtn5h3AwPxrXS5R7czBZQo6homDf8AfJGTU1FAHHW9vPF4V0W3NpcLLBexvKggbKhX3Menofxq9DYyQXWrRXn2por6YyoYo9wkRkVdpODtIxjkgYx710dFAEVtH5NpDEAV2Iq4LbsYHr3+tZV4hbxdYytBK8UdtLG0giYqrOUxzjHQH+tbVFAGAlpLZ6/qM0zXAguxGYmgi3gbV2lCNpI9R25piW32TWdFjhtrhYLW1lhJ2M4j3eXtBYDH8J+mK6KigDmLDTHh0mXS78XUj+e7YRfklzIXVt+OOozk5BH0q5p0f/FS6tJJbyBZjD5cjwsFbYuDyR2P/wBatuigDmLHTJIdKm0y/F1K5nkbCL8k26Qurb8cdRnJyCPpU8QaPXdcma3n2ywxLG/kMd+1WBA455IroKKAOU061ubeLw9dTQTFLOzNtPF5Z3ROVUbtvUj5SOPX0zVq/snfULC/tIJIbeCeVpVijAdt6Y8zaRknOQeM4Oa6GigCjp1ukc11PGs2Z3VmeYbS5CgZ24GOAB05xUHieOSfwxfwQRvLLNCyIkaliSRx0rVooAwZ90niHSJIYJxFFbzI7eQwCFtm0Hjj7p+mKraVprw6NbaZfi6ea3lUlVXCMVk3Bw+OnG7rntXT0UAYunpcWmtakrQPtubpZVfaduzylBOemcrjHWs650m+VtRt7aNhFbzfb7JgcBpT82z6bg+R/tiurooAwbuzuZYtNvfKctHd/ariDGWwVIHHcoCOP9njmieOWfWp9QjjmW3SwaDBiYNK5bIwuM8Y64/i+tb1FAHO28DR/D8WwtpVuV0/yGj8lt2/y8Yxjn5j16Umj/adOvYY5o7m4huoEPnPCd0DqAuxuBhT1Hpz65ro6KACuGk0k3MWqX66bcrfQ6gtzbs6EFkDLnA6NwG49xXc0gIZQVOQeQR3oA5G8i1GTV4dY8vULWOe0WMpbRxySQEMThlZScEEdO45p0kE1n4FuLGOxvnluo51jj2K7KXLEbtoAXOc4HTpXWZAIGeT0FLQBg388l54LvUisrsSvbPAsLREOWKYHHpk9frVe0tLlfFVvcJbSrZywvOS648uRlVWUjsTtB+pNdNRQBwcej6hDZHSrn+05W+0bwIo4RDJ8+4SeYVLKe5zz2roNVM9n4gs9RS0nuoVtpYWFuu5gzFCOPQ7TzW5RQBDaPPLZQvdxCGdkBkjVtwRscjPeqXiG0lutHka0XddW7LcW4Hd0OQPx5H41p0UAc9a2N5Hq+Xib7PfqtzcZOfKkQ/d/EFB9ENWVD/8Jm0vlS+V9iEXmeW23dvLYzjHStiigDD8P+fYaYtrc28q+QZjIfLY8mQlduPvZBJ4zWXa288PhfQLc2lwsttcxPMggbKBSdxPHv8AjXYUUAFZH2NpfFq3DQkQ21sSjkcNLI2CQfUKgH/Aq16KAOasp73TdAuLp7bbcPOZpEmBALPJyo+gKjPIrQl1C9jvkshDG05ieYuoJUgNhQASOeRnnj3zWjNBDcxmO4iSVCclZFDD8jSSWsEyqs0Ecip90MgIX6UAZEQe/wDEMT3KoEWzjmWI/MY2LHOGBxnIHI6gVYsh5XiPU0brKkMy+42lf5r+oq+LWBbjz1gjE23b5gQbsemfSkltkkuIp8lZYsgMO6nqp9jgfkKAMB7a8/s3X7qO3YXd5K0cQK/MI1AjU46kcM3vn3rTjuf7Ps7GEWkxiZ1t1IH+rXorNnpnj860qY8SyFC4zsbcBnjNAGfrcRureO1eOYwSuDJNAxDwlfmVlABJO4D2pNDt57a3uGu5DKXmLJNJGEkdAoAL4A54I+mK1Khurdbu3aCQsI34cLxuXuPx6UAUfDikaBA5BAmaSZQR/C7s4/RhWpSABVAUYA4AHaloAKKKKACiiigAooooAKKKKACiiigCC9uTaWU1wsTzGJC2xCAWx9SBWfa6yy+GItUvbebC2yzSlQpLDblmAz0/L6VpXMX2i0mhBwZEZc+mRisT7BqLeEJtMaEeb9gFoi71wX2FS+f7vT346UAXv7dtRbSzuk6RR+Xh3jwJN5wu09DyR9O+Kmj1CO4a6gWOZLi3UF4uA2GB2kHOOcHv2qreQahLokEVqGglVoxMgkAYoCNwVh0JA4P8qgtLC7s9euLyG1C288MalTIC/wAm/rz1JZecngHNADrXWbSy0LT5y15PBOViSabDyZJwN/ckn0zV+LU4ZLi4gdZIZbdFkkSQD7pzhhgnP3T+VYcelagnhzS7M2uZrS6jlkAkXG1H3cHPerklpqJ1rUbq2jWIz2awwSOwIWRd5BIHbLD8qALlrrNvdS2yKsifaozLbs4GJVGCcYJwcEHBxWdpF19kudaQRXEscN6T8vzbF8tDxk5Pc4GfpyKjGl34utMuooD51srrM8829mZkxuznlQe3HXgVdtba90+51ORLdZvtdx50WJAAPkVfmz0Hy9s/SgCK7mjn17Qrm1l3xz+bhkY7XXyyRx0NXDrVqFaQCQ26zeS1wANgfdtx1zjdxnGM1m22i3OnHQoIUM8enh/Ol3AZLKRwCfU5+lPsNENlYtp0lolygkYxzO2UKs5YblznIz0A5wORngA32JVSQpYgZwOprKi8RWsmnm+MNxHaq7JJK6ACMhipzznAI6jIH51rVyemw3d94PubC3hUfaZbmITMw2qrSuCSOuRk4GPTkdgDdudXgt5pYgkkrQQiaURgHahzg8kZ+6emelJca1bWxsiVkkjviFgkjAKsSpYDrnkCqEmlXfnTwOGuLX7MkNoDJhYyFwxde5zg5wePTvDHp+pJZaHE9qrHT3QuFkX7qwlOMnk7iT9PegDVXWIX+zqkFx51wrukDIEfCkBs7iB3HfvTJtes4dITU8SPaswVmVeUJbbyDzw3BpL63u5dWgk2vNYrEwMMcmwmTIwW5G5cZGM9ex7ZH9j6mnheXSxbxs32jehSQAEeeZCee2MDp1z2oA3P7Xi+2/YzDOLkx+akRUZdM4yDnAx746j1pF1q0ewiulLkTSmGOPb87SAkFceoIPtwTnFQGC7PihL/AOyt5AszCfnXO4sG6Z9sfX25rNg0K7NjamaECezvpbkRmQYkR2fIyOhw35j8aAL2r699k0fUJreCX7TaLhoztzGWHysecEfTPTFaE2ox21uklxFJG8kgjjiOCzsegGCR69+xzWdqGkPf6PqUNvbR2st1CEUucsxGSNxBPHPHXHNLqGnXWpfZbp49slrc+alu7AZQptZSRxnkkH6dOwBYudetrS1uZp4pl+yEefHhdyA9G64I9wT+hp66xC0nliG4EjSCONWTb5nyb8qScEY/Lp1qhqumTXejahFZ2KxTXcaxjcw3HGeWOSMDPAGep9atapZnU4LaO5sFng3lponYB04OCpB659xxmgC/bXIuVk/dyRtG5RlkAznAPYn1qasSCO80fS38iMzJ9pBjhuJizRxHAK7ucnOcDPcDPFbdAHHpdWlhrXiI3z3DQ2vlPHH5kjBR5QYgYPGTW6og0KC5lnmf7PJKGjVnaRgxAXYuck5I4A9apWdreQ6xrFzcae7wX7R7UEiE4WPaQee9Le6fqGqf6RJEsEltdRTW0LSZDhMk7iOhO5h3xgUAacWpxPfJZyRyw3DxtKI3A+6CBnIJHUjvVysEyyyeNrMyQmEfYJsBmBY/PHnpkfrWvZvdPbBr6KOKbc2VjfcMZOOcDtigCpqshkubKwU4W6kJlwesaLkj8TtB9iar6vbzT6lAsV7f2qJETtswCG5HXII6dKm1JfK1bS7sj5FkeFz/AHd68H/vpVH/AAKppRdwai86Ca4geIKsCFAEYE5POOoI79qAIJwl14d87TZ2laOMy28zMSxdfUnnkggj0JFX7S4W7soLmP7k0ayL9CM1kWttJoXhW488h5z5spRDxvdiQi/iQB6mrtrBdWFhptnbxxyJEiRTsz4KqqYyOOTkCgDQooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigAooooAKKKKACiiigBGUOpVhlSMEHvUNrZ21lGUs7eOBCclY0Cgn14qeigAooooAKKKKACiiigAooooAKKKKAEZFfG9Q2DkZHQ+tLRRQAUUUUAV2sbR7xbp7aJrhRhZSg3AemetWKKKAGSxJPE0cqhkYYIPenKNqgDJwMcnNLRQBHJDHKyNIu7y23KCeAfXFSUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAFFFFABRRRQAUUUUAf/9k=\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 5,
     "metadata": {
      "image/jpeg": {
       "width": 400
      }
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.display import Image\n",
    "Image(filename=\"C:/Users/sneha/Capture.JPG\", width=400)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data Processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "size of input is  torch.Size([441, 2])\n",
      "size of output is  torch.Size([441, 6129])\n",
      "   NODE             X          Y         Z     A      B  Number Experiment\n",
      "0     1 -1.060000e-14  40.048313  0.000021 -0.01 -0.001                  1\n",
      "1     2  1.582933e+00  39.936877 -0.000020 -0.01 -0.001                  1\n",
      "2     3  3.137810e+00  39.648188 -0.000137 -0.01 -0.001                  1\n",
      "3     4  4.728823e+00  40.078871  0.000057 -0.01 -0.001                  1\n",
      "4     5  6.312642e+00  39.612751 -0.000137 -0.01 -0.001                  1\n",
      "torch.Size([441, 6129])\n",
      "the train size is 400 and validation size is 41 \n"
     ]
    }
   ],
   "source": [
    "from torch.utils.data import Dataset, DataLoader\n",
    "   \n",
    "\n",
    "\n",
    "class Dataset(Dataset):\n",
    "    def __init__(self, transform=None):\n",
    "        super(Dataset,self).__init__()\n",
    "        self.data_name=pd.read_csv('D:/JOB, Application STUFF/Fraunhofer/Frauhofer task/Takk 3/Example_Data.csv', sep=';', decimal=',')\n",
    "        self.x = torch.tensor(self.data_name[['A','B']].values.reshape(441,-1,2)[:,1,:].astype('float32'))\n",
    "        print(\"size of input is \",self.x.size())\n",
    "        #X, Y, Z are outputs\n",
    "        X = self.data_name['X'].values.reshape(441,2043)\n",
    "        Y = self.data_name['Y'].values.reshape(441,2043)\n",
    "        Z = self.data_name['Z'].values.reshape(441,2043)\n",
    "        self.y = torch.from_numpy(np.concatenate((X,Y,Z),axis=1).astype('float32'))\n",
    "        \n",
    "        print(\"size of output is \",self.y.size())\n",
    "        self.transform=transform \n",
    "        print(self.data_name.head())\n",
    "        self.len = self.x.shape[0]\n",
    "        \n",
    "    def __len__(self):\n",
    "        return self.len\n",
    "     \n",
    "    def __getitem__(self, index):\n",
    "        return self.x[index],self.y[index]\n",
    "    \n",
    "data = Dataset()\n",
    "\n",
    "#print(\"size of the data is \",len(data))\n",
    "sample = data[:][1]\n",
    "print(sample.shape)\n",
    "\n",
    "#splitting data into train and vaidation\n",
    "\n",
    "val_size = 41\n",
    "train_size = len(data) - val_size\n",
    "train_ds, val_ds = random_split(data, [train_size, val_size])\n",
    "print(\"the train size is {} and validation size is {} \".format(len(train_ds), len(val_ds)))\n",
    "\n",
    "batch_size = 100\n",
    "train_loader = DataLoader(train_ds, batch_size,num_workers=8 shuffle=True)\n",
    "val_loader = DataLoader(val_ds, batch_size, num_workers=8) \n",
    "\n",
    "\n",
    "#for x,y in train_loader:\n",
    "    #print(x.shape)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Creating Pruning matrix and Distance Matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [],
   "source": [
    "D = np.zeros((m,m))\n",
    "for i in range (m):\n",
    "    for j in range (m):\n",
    "        D[i,j] = np.sqrt((((xi[i]-xi[j])**2)+(yi[i]-yi[j])**2))\n",
    "#print(D)   "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Calculate the hidden size for Refnet by Cupnet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [],
   "source": [
    "def hidden(alpha, h, d):\n",
    "    C = np.ones((m,m))\n",
    "    for i in range(m):\n",
    "        for j in range(m):\n",
    "            if(D[i,j]>alpha):\n",
    "                C[i,j]=0\n",
    "    #C = torch.tensor(C)\n",
    "    #print(C)\n",
    "    c_alpha_sum = C.sum()\n",
    "    u = in_size + h + d + 1\n",
    "    n_cup = ((in_size*d + d)+  3*h*(int(c_alpha_sum) + m) + 3*(m**2 + m))\n",
    "    s = int((-u + math.sqrt(u**2 - 4*h*(d-(int(n_cup)))/(2*h))))\n",
    "    n_ref = (((in_size*s) + s) + h*(s**2 + s) + (s*d + d)) \n",
    "    #print(\"Total trainable parameter in Ref net is \", n_ref)\n",
    "    #print(\"Total tainable paramter in cup net is \",n_cup)\n",
    "    print(\"Hidden layer size of Refnet is \",s)\n",
    "    return C,s\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Defining the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "class My_Model(nn.Module):\n",
    "    def __init__(self, in_size, alpha, h, d, p=0):\n",
    "        \n",
    "        super(My_Model,self). __init__()\n",
    "        self.in_size = in_size\n",
    "        self.alpha = alpha\n",
    "        self.h = h\n",
    "        self.d = d\n",
    "        #calling hidden function\n",
    "        self.hidden_obj,self.s = hidden(self.alpha, self.h,self.d)\n",
    "        self.hidden=nn.ModuleList()\n",
    "        self.drop=nn.Dropout(p=p)\n",
    "        self.Layers = [in_size, *[self.s for i in range(h)], 6129]\n",
    "        print(self.Layers)\n",
    "        for input_size,output_size in zip(self.Layers, self.Layers[1:]):\n",
    "            self.hidden.append(nn.Linear(input_size, output_size))\n",
    "            \n",
    "    def forward(self, activation):\n",
    "        L=len(self.hidden)\n",
    "        for(l,linear_transform) in zip(range(L),self.hidden):\n",
    "            if l<L-1:\n",
    "                activation=torch.relu(linear_transform(activation))\n",
    "                activation=self.drop(activation)\n",
    "            else:\n",
    "                activation= linear_transform(activation)\n",
    "        return activation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import r2_score\n",
    "\n",
    "#accesing the weights of the optimised model\n",
    "#print(model.parameters())\n",
    "#print(model.state_dict()['hidden.0.weight'])#[hidden.weight]\n",
    "#print(model.state_dict()['hidden.1.weight'].shape)\n",
    "#model.state_dict()['hidden.1.weight'] = model.state_dict()['hidden.1.weight']*hidden_obj\n",
    "\n",
    "\n",
    "def train(model,train_loader, val_loader, optimizer, epochs=20):\n",
    "    total_loss = {'training_loss':[],'validation_loss':[]}\n",
    "    for epoch in range(epochs):\n",
    "        batch_loss=[]\n",
    "        for x,y in train_loader:\n",
    "            optimizer.zero_grad()\n",
    "            yhat = model(x) \n",
    "            loss = MSE(y,yhat)\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            batch_loss.append(loss.item())  \n",
    "        total_loss['training_loss'].append(torch.Tensor(batch_loss).mean().item())\n",
    "    \n",
    "        batch_loss=[]        \n",
    "        for x,y in val_loader:\n",
    "            optimizer.zero_grad()\n",
    "            yhat = model(x) \n",
    "            loss = MSE(y,yhat)\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            batch_loss.append(loss.item())\n",
    "            total_loss['validation_loss'].append(torch.Tensor(batch_loss).mean().item()) \n",
    "        if total_loss['training_loss'][epoch]>total_loss['training_loss'][epoch-1] and total_loss['validation_loss'][epoch]>total_loss['validation_loss'][epoch-1]:\n",
    "            break\n",
    "    return total_loss\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### single model loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 211,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:178.33,     validation_loss:27.23\n",
      "training_loss:99.56,     validation_loss:4.20\n",
      "training_loss:89.08,     validation_loss:2.70\n",
      "training_loss:87.80,     validation_loss:1.52\n",
      "training_loss:87.60,     validation_loss:1.93\n",
      "training_loss:87.60,     validation_loss:1.79\n",
      "training_loss:87.80,     validation_loss:1.76\n",
      "training_loss:87.80,     validation_loss:1.96\n"
     ]
    }
   ],
   "source": [
    "in_size = 2 \n",
    "d = 2043*3 \n",
    "alpha = 1\n",
    "h = 1\n",
    "torch.manual_seed(55)\n",
    "\n",
    "MSE = nn.MSELoss()\n",
    "model = My_Model(in_size, alpha, h,d, p=0.20)\n",
    "optimizer = optim.Adam(model.parameters(), lr=5E-2, weight_decay= 1e-3)\n",
    "total_loss= train(model,train_loader, val_loader, optimizer)\n",
    "\n",
    "for training_loss,validation_loss in zip(total_loss['training_loss'],total_loss['validation_loss']):\n",
    "    print('training_loss:{:.2f},     validation_loss:{:.2f}'.format(training_loss,validation_loss))\n",
    "\n",
    "    \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### to find best learning rate\n",
    "https://github.com/davidtvs/pytorch-lr-finder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "02d43a28291a4b20909d75a5b8bc2b7f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Stopping early, the loss has diverged\n",
      "Learning rate search finished. See the graph with {finder_name}.plot()\n",
      "LR suggestion: steepest gradient\n",
      "Suggested LR: 8.11E-07\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from torch_lr_finder import LRFinder\n",
    "\n",
    "#def LRfinder():\n",
    "optimizer = optim.Adam(model.parameters(), lr=1e-7, weight_decay=1e-2)\n",
    "lr_finder = LRFinder(model, optimizer, MSE, device=\"cuda\")\n",
    "lr_finder.range_test(train_loader, end_lr=100, num_iter=100)\n",
    "lr_finder.plot() # to inspect the loss-learning rate graph\n",
    "lr_finder.reset() # to reset the model and optimizer to their initial state\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "random seeds are = [76684 86728 48351 19514 68167 48275 30650 54546  4927 51683]\n",
      "alpha = 1, hidden layer = 1 seed = 76684\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:112.11,     validation_loss:25.64\n",
      "training_loss:111.81,     validation_loss:25.15\n",
      "training_loss:111.67,     validation_loss:24.93\n",
      "training_loss:111.41,     validation_loss:25.23\n",
      "training_loss:110.91,     validation_loss:24.17\n",
      "training_loss:110.89,     validation_loss:23.99\n",
      "training_loss:110.65,     validation_loss:24.37\n",
      "training_loss:110.13,     validation_loss:23.79\n",
      "training_loss:110.06,     validation_loss:23.02\n",
      "training_loss:109.75,     validation_loss:23.01\n",
      "training_loss:109.60,     validation_loss:23.01\n",
      "training_loss:109.20,     validation_loss:22.35\n",
      "training_loss:109.13,     validation_loss:22.70\n",
      "training_loss:108.80,     validation_loss:22.43\n",
      "training_loss:108.53,     validation_loss:21.64\n",
      "training_loss:108.50,     validation_loss:22.02\n",
      "training_loss:108.26,     validation_loss:21.36\n",
      "training_loss:107.85,     validation_loss:21.66\n",
      "training_loss:107.74,     validation_loss:21.37\n",
      "training_loss:107.53,     validation_loss:21.43\n",
      "alpha = 1, hidden layer = 1 seed = 86728\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:107.47,     validation_loss:20.99\n",
      "training_loss:107.09,     validation_loss:20.79\n",
      "training_loss:106.82,     validation_loss:20.62\n",
      "training_loss:106.64,     validation_loss:20.02\n",
      "training_loss:106.49,     validation_loss:20.03\n",
      "training_loss:106.24,     validation_loss:19.90\n",
      "training_loss:105.98,     validation_loss:20.27\n",
      "training_loss:105.93,     validation_loss:19.54\n",
      "training_loss:105.74,     validation_loss:18.71\n",
      "training_loss:105.57,     validation_loss:18.86\n",
      "training_loss:105.45,     validation_loss:18.85\n",
      "training_loss:104.94,     validation_loss:18.51\n",
      "training_loss:105.09,     validation_loss:18.81\n",
      "alpha = 1, hidden layer = 1 seed = 48351\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:104.85,     validation_loss:18.77\n",
      "training_loss:104.59,     validation_loss:18.27\n",
      "training_loss:104.60,     validation_loss:18.39\n",
      "alpha = 1, hidden layer = 1 seed = 19514\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:104.29,     validation_loss:17.97\n",
      "training_loss:103.93,     validation_loss:17.60\n",
      "training_loss:104.15,     validation_loss:17.27\n",
      "training_loss:103.86,     validation_loss:17.14\n",
      "training_loss:103.66,     validation_loss:17.22\n",
      "training_loss:103.66,     validation_loss:16.66\n",
      "training_loss:103.32,     validation_loss:16.96\n",
      "training_loss:103.35,     validation_loss:17.06\n",
      "alpha = 1, hidden layer = 1 seed = 68167\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:103.27,     validation_loss:16.93\n",
      "training_loss:103.01,     validation_loss:16.65\n",
      "training_loss:103.20,     validation_loss:15.97\n",
      "training_loss:103.04,     validation_loss:16.57\n",
      "training_loss:102.70,     validation_loss:16.54\n",
      "training_loss:102.96,     validation_loss:16.21\n",
      "training_loss:102.85,     validation_loss:16.55\n",
      "training_loss:102.83,     validation_loss:16.71\n",
      "training_loss:102.63,     validation_loss:16.15\n",
      "training_loss:102.80,     validation_loss:16.45\n",
      "alpha = 1, hidden layer = 1 seed = 48275\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:102.72,     validation_loss:15.93\n",
      "training_loss:102.69,     validation_loss:16.61\n",
      "training_loss:102.59,     validation_loss:16.46\n",
      "training_loss:102.67,     validation_loss:16.46\n",
      "training_loss:102.63,     validation_loss:15.88\n",
      "training_loss:102.71,     validation_loss:16.29\n",
      "alpha = 1, hidden layer = 1 seed = 30650\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:102.63,     validation_loss:16.10\n",
      "training_loss:102.72,     validation_loss:16.39\n",
      "alpha = 1, hidden layer = 1 seed = 54546\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:102.81,     validation_loss:16.01\n",
      "training_loss:102.50,     validation_loss:16.34\n",
      "training_loss:102.62,     validation_loss:16.41\n",
      "alpha = 1, hidden layer = 1 seed = 4927\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:102.73,     validation_loss:15.91\n",
      "training_loss:102.67,     validation_loss:15.72\n",
      "training_loss:102.72,     validation_loss:16.44\n",
      "alpha = 1, hidden layer = 1 seed = 51683\n",
      "Hidden layer size of Refnet is  1790\n",
      "[2, 1790, 6129]\n",
      "training_loss:102.64,     validation_loss:16.28\n",
      "training_loss:102.64,     validation_loss:16.12\n",
      "training_loss:102.63,     validation_loss:16.39\n",
      "training_loss:102.46,     validation_loss:16.55\n",
      "training_loss:102.51,     validation_loss:16.23\n",
      "training_loss:102.72,     validation_loss:16.27\n",
      "alpha = 1, hidden layer = 2 seed = 76684\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.67,     validation_loss:16.32\n",
      "training_loss:102.69,     validation_loss:15.91\n",
      "training_loss:102.67,     validation_loss:16.24\n",
      "training_loss:102.57,     validation_loss:16.13\n",
      "training_loss:102.67,     validation_loss:16.15\n",
      "alpha = 1, hidden layer = 2 seed = 86728\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.55,     validation_loss:16.29\n",
      "training_loss:102.64,     validation_loss:16.36\n",
      "alpha = 1, hidden layer = 2 seed = 48351\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.66,     validation_loss:16.09\n",
      "training_loss:102.66,     validation_loss:16.03\n",
      "training_loss:102.57,     validation_loss:16.43\n",
      "training_loss:102.62,     validation_loss:16.41\n",
      "training_loss:102.57,     validation_loss:16.45\n",
      "training_loss:102.50,     validation_loss:16.38\n",
      "training_loss:102.63,     validation_loss:16.07\n",
      "training_loss:102.64,     validation_loss:15.91\n",
      "training_loss:102.70,     validation_loss:16.19\n",
      "alpha = 1, hidden layer = 2 seed = 19514\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.66,     validation_loss:16.38\n",
      "training_loss:102.50,     validation_loss:16.43\n",
      "training_loss:102.64,     validation_loss:15.73\n",
      "training_loss:102.72,     validation_loss:16.43\n",
      "alpha = 1, hidden layer = 2 seed = 68167\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.61,     validation_loss:16.29\n",
      "training_loss:102.58,     validation_loss:16.46\n",
      "training_loss:102.60,     validation_loss:15.95\n",
      "training_loss:102.55,     validation_loss:16.69\n",
      "training_loss:102.57,     validation_loss:15.96\n",
      "training_loss:102.64,     validation_loss:16.27\n",
      "alpha = 1, hidden layer = 2 seed = 48275\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.76,     validation_loss:16.13\n",
      "training_loss:102.54,     validation_loss:15.84\n",
      "training_loss:102.83,     validation_loss:15.95\n",
      "alpha = 1, hidden layer = 2 seed = 30650\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.68,     validation_loss:16.16\n",
      "training_loss:102.49,     validation_loss:16.17\n",
      "training_loss:102.70,     validation_loss:16.64\n",
      "alpha = 1, hidden layer = 2 seed = 54546\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.67,     validation_loss:16.20\n",
      "training_loss:102.56,     validation_loss:16.05\n",
      "training_loss:102.87,     validation_loss:15.98\n",
      "training_loss:102.43,     validation_loss:16.49\n",
      "training_loss:102.73,     validation_loss:16.18\n",
      "training_loss:102.77,     validation_loss:16.38\n",
      "alpha = 1, hidden layer = 2 seed = 4927\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.55,     validation_loss:15.96\n",
      "training_loss:102.76,     validation_loss:16.01\n",
      "alpha = 1, hidden layer = 2 seed = 51683\n",
      "Hidden layer size of Refnet is  1796\n",
      "[2, 1796, 1796, 6129]\n",
      "training_loss:102.70,     validation_loss:16.01\n",
      "training_loss:102.58,     validation_loss:16.72\n",
      "training_loss:102.57,     validation_loss:16.11\n",
      "training_loss:102.53,     validation_loss:16.16\n",
      "training_loss:102.63,     validation_loss:16.26\n",
      "alpha = 1, hidden layer = 3 seed = 76684\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.72,     validation_loss:15.89\n",
      "training_loss:102.66,     validation_loss:16.36\n",
      "training_loss:102.66,     validation_loss:16.61\n",
      "alpha = 1, hidden layer = 3 seed = 86728\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.65,     validation_loss:16.10\n",
      "training_loss:102.68,     validation_loss:16.37\n",
      "alpha = 1, hidden layer = 3 seed = 48351\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.65,     validation_loss:16.48\n",
      "training_loss:102.56,     validation_loss:16.30\n",
      "training_loss:102.58,     validation_loss:16.19\n",
      "training_loss:102.59,     validation_loss:16.38\n",
      "alpha = 1, hidden layer = 3 seed = 19514\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.69,     validation_loss:16.61\n",
      "training_loss:102.55,     validation_loss:16.23\n",
      "training_loss:102.52,     validation_loss:16.08\n",
      "training_loss:102.62,     validation_loss:16.05\n",
      "training_loss:102.60,     validation_loss:16.03\n",
      "training_loss:102.64,     validation_loss:16.19\n",
      "alpha = 1, hidden layer = 3 seed = 68167\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.61,     validation_loss:16.25\n",
      "training_loss:102.84,     validation_loss:15.86\n",
      "training_loss:102.59,     validation_loss:16.18\n",
      "training_loss:102.59,     validation_loss:15.92\n",
      "training_loss:102.74,     validation_loss:16.19\n",
      "alpha = 1, hidden layer = 3 seed = 48275\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.63,     validation_loss:16.30\n",
      "training_loss:102.67,     validation_loss:16.26\n",
      "training_loss:102.55,     validation_loss:16.56\n",
      "training_loss:102.69,     validation_loss:16.20\n",
      "training_loss:102.51,     validation_loss:16.62\n",
      "training_loss:102.67,     validation_loss:16.11\n",
      "training_loss:102.54,     validation_loss:16.13\n",
      "training_loss:102.61,     validation_loss:16.37\n",
      "alpha = 1, hidden layer = 3 seed = 30650\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.74,     validation_loss:16.47\n",
      "training_loss:102.57,     validation_loss:16.61\n",
      "training_loss:102.67,     validation_loss:15.80\n",
      "training_loss:102.61,     validation_loss:16.09\n",
      "training_loss:102.87,     validation_loss:16.03\n",
      "training_loss:102.51,     validation_loss:16.24\n",
      "training_loss:102.59,     validation_loss:16.04\n",
      "training_loss:102.64,     validation_loss:16.40\n",
      "alpha = 1, hidden layer = 3 seed = 54546\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.76,     validation_loss:16.27\n",
      "training_loss:102.59,     validation_loss:15.86\n",
      "training_loss:102.67,     validation_loss:15.91\n",
      "alpha = 1, hidden layer = 3 seed = 4927\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.66,     validation_loss:16.41\n",
      "training_loss:102.66,     validation_loss:16.42\n",
      "training_loss:102.58,     validation_loss:15.89\n",
      "training_loss:102.70,     validation_loss:16.35\n",
      "alpha = 1, hidden layer = 3 seed = 51683\n",
      "Hidden layer size of Refnet is  1802\n",
      "[2, 1802, 1802, 1802, 6129]\n",
      "training_loss:102.70,     validation_loss:16.54\n",
      "training_loss:102.62,     validation_loss:16.09\n",
      "training_loss:102.58,     validation_loss:16.16\n",
      "training_loss:102.64,     validation_loss:16.15\n",
      "training_loss:102.51,     validation_loss:16.07\n",
      "training_loss:102.78,     validation_loss:16.21\n",
      "alpha = 1, hidden layer = 4 seed = 76684\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.74,     validation_loss:16.38\n",
      "training_loss:102.59,     validation_loss:16.30\n",
      "training_loss:102.64,     validation_loss:16.28\n",
      "training_loss:102.72,     validation_loss:16.22\n",
      "training_loss:102.47,     validation_loss:16.16\n",
      "training_loss:102.66,     validation_loss:16.53\n",
      "alpha = 1, hidden layer = 4 seed = 86728\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.82,     validation_loss:15.94\n",
      "training_loss:102.62,     validation_loss:16.29\n",
      "training_loss:102.81,     validation_loss:15.82\n",
      "training_loss:102.68,     validation_loss:16.19\n",
      "training_loss:102.72,     validation_loss:16.09\n",
      "training_loss:102.59,     validation_loss:16.13\n",
      "training_loss:102.73,     validation_loss:16.23\n",
      "alpha = 1, hidden layer = 4 seed = 48351\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.71,     validation_loss:16.59\n",
      "training_loss:102.56,     validation_loss:16.27\n",
      "training_loss:102.67,     validation_loss:15.73\n",
      "training_loss:102.69,     validation_loss:16.26\n",
      "alpha = 1, hidden layer = 4 seed = 19514\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.66,     validation_loss:16.58\n",
      "training_loss:102.46,     validation_loss:16.81\n",
      "training_loss:102.53,     validation_loss:16.29\n",
      "training_loss:102.69,     validation_loss:16.03\n",
      "training_loss:102.54,     validation_loss:16.33\n",
      "training_loss:102.73,     validation_loss:16.12\n",
      "training_loss:102.55,     validation_loss:15.99\n",
      "training_loss:102.64,     validation_loss:16.42\n",
      "alpha = 1, hidden layer = 4 seed = 68167\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.66,     validation_loss:16.34\n",
      "training_loss:102.62,     validation_loss:16.36\n",
      "training_loss:102.68,     validation_loss:16.44\n",
      "alpha = 1, hidden layer = 4 seed = 48275\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.68,     validation_loss:16.01\n",
      "training_loss:102.75,     validation_loss:16.34\n",
      "alpha = 1, hidden layer = 4 seed = 30650\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.68,     validation_loss:16.55\n",
      "training_loss:102.39,     validation_loss:16.46\n",
      "training_loss:102.73,     validation_loss:16.04\n",
      "training_loss:102.63,     validation_loss:15.94\n",
      "training_loss:102.62,     validation_loss:16.32\n",
      "training_loss:102.65,     validation_loss:16.41\n",
      "alpha = 1, hidden layer = 4 seed = 54546\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.60,     validation_loss:16.30\n",
      "training_loss:102.58,     validation_loss:16.27\n",
      "training_loss:102.63,     validation_loss:15.96\n",
      "training_loss:102.72,     validation_loss:16.04\n",
      "alpha = 1, hidden layer = 4 seed = 4927\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.65,     validation_loss:16.10\n",
      "training_loss:102.58,     validation_loss:16.09\n",
      "training_loss:102.70,     validation_loss:16.44\n",
      "alpha = 1, hidden layer = 4 seed = 51683\n",
      "Hidden layer size of Refnet is  1808\n",
      "[2, 1808, 1808, 1808, 1808, 6129]\n",
      "training_loss:102.61,     validation_loss:16.20\n",
      "training_loss:102.65,     validation_loss:16.09\n",
      "training_loss:102.64,     validation_loss:16.12\n",
      "training_loss:102.72,     validation_loss:16.22\n",
      "alpha = 1, hidden layer = 5 seed = 76684\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n",
      "training_loss:102.64,     validation_loss:16.20\n",
      "training_loss:102.59,     validation_loss:16.01\n",
      "training_loss:102.68,     validation_loss:16.16\n",
      "alpha = 1, hidden layer = 5 seed = 86728\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n",
      "training_loss:102.64,     validation_loss:16.38\n",
      "training_loss:102.58,     validation_loss:16.32\n",
      "training_loss:102.72,     validation_loss:15.95\n",
      "training_loss:102.63,     validation_loss:16.39\n",
      "training_loss:102.58,     validation_loss:16.40\n",
      "training_loss:102.68,     validation_loss:16.34\n",
      "training_loss:102.64,     validation_loss:16.33\n",
      "training_loss:102.51,     validation_loss:16.60\n",
      "training_loss:102.57,     validation_loss:16.27\n",
      "training_loss:102.69,     validation_loss:16.10\n",
      "training_loss:102.60,     validation_loss:16.38\n",
      "training_loss:102.55,     validation_loss:16.37\n",
      "training_loss:102.71,     validation_loss:15.84\n",
      "training_loss:102.61,     validation_loss:16.14\n",
      "training_loss:102.53,     validation_loss:16.16\n",
      "training_loss:102.60,     validation_loss:16.14\n",
      "training_loss:102.58,     validation_loss:16.17\n",
      "training_loss:102.65,     validation_loss:16.12\n",
      "training_loss:102.48,     validation_loss:16.34\n",
      "training_loss:102.69,     validation_loss:16.15\n",
      "alpha = 1, hidden layer = 5 seed = 48351\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n",
      "training_loss:102.65,     validation_loss:16.20\n",
      "training_loss:102.60,     validation_loss:16.19\n",
      "training_loss:102.59,     validation_loss:16.48\n",
      "training_loss:102.59,     validation_loss:15.87\n",
      "training_loss:102.66,     validation_loss:15.99\n",
      "alpha = 1, hidden layer = 5 seed = 19514\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training_loss:102.70,     validation_loss:16.20\n",
      "training_loss:102.79,     validation_loss:16.33\n",
      "alpha = 1, hidden layer = 5 seed = 68167\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n",
      "training_loss:102.70,     validation_loss:16.13\n",
      "training_loss:102.69,     validation_loss:16.18\n",
      "training_loss:102.64,     validation_loss:15.98\n",
      "training_loss:102.63,     validation_loss:16.17\n",
      "training_loss:102.65,     validation_loss:15.92\n",
      "training_loss:102.54,     validation_loss:16.39\n",
      "training_loss:102.60,     validation_loss:16.15\n",
      "training_loss:102.67,     validation_loss:16.00\n",
      "training_loss:102.68,     validation_loss:16.32\n",
      "alpha = 1, hidden layer = 5 seed = 48275\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n",
      "training_loss:102.65,     validation_loss:16.13\n",
      "training_loss:102.67,     validation_loss:16.23\n",
      "alpha = 1, hidden layer = 5 seed = 30650\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n",
      "training_loss:102.71,     validation_loss:16.27\n",
      "training_loss:102.53,     validation_loss:16.16\n",
      "training_loss:102.68,     validation_loss:15.87\n",
      "training_loss:102.67,     validation_loss:16.38\n",
      "training_loss:102.61,     validation_loss:15.79\n",
      "training_loss:102.57,     validation_loss:16.44\n",
      "training_loss:102.54,     validation_loss:15.90\n",
      "training_loss:102.62,     validation_loss:15.78\n",
      "training_loss:102.73,     validation_loss:15.96\n",
      "alpha = 1, hidden layer = 5 seed = 54546\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n",
      "training_loss:102.67,     validation_loss:16.05\n",
      "training_loss:102.74,     validation_loss:16.02\n",
      "training_loss:102.61,     validation_loss:16.25\n",
      "training_loss:102.73,     validation_loss:16.01\n",
      "training_loss:102.64,     validation_loss:16.02\n",
      "training_loss:102.75,     validation_loss:16.16\n",
      "alpha = 1, hidden layer = 5 seed = 4927\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n",
      "training_loss:102.71,     validation_loss:16.27\n",
      "training_loss:102.52,     validation_loss:16.26\n",
      "training_loss:102.68,     validation_loss:16.41\n",
      "alpha = 1, hidden layer = 5 seed = 51683\n",
      "Hidden layer size of Refnet is  1813\n",
      "[2, 1813, 1813, 1813, 1813, 1813, 6129]\n",
      "training_loss:102.59,     validation_loss:16.44\n",
      "training_loss:102.60,     validation_loss:16.06\n",
      "training_loss:102.62,     validation_loss:16.51\n",
      "alpha = 1, hidden layer = 6 seed = 76684\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.60,     validation_loss:15.91\n",
      "training_loss:102.79,     validation_loss:16.07\n",
      "alpha = 1, hidden layer = 6 seed = 86728\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.75,     validation_loss:15.95\n",
      "training_loss:102.58,     validation_loss:16.17\n",
      "training_loss:102.65,     validation_loss:16.06\n",
      "training_loss:102.72,     validation_loss:16.31\n",
      "alpha = 1, hidden layer = 6 seed = 48351\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.71,     validation_loss:16.26\n",
      "training_loss:102.65,     validation_loss:15.99\n",
      "training_loss:102.78,     validation_loss:16.15\n",
      "alpha = 1, hidden layer = 6 seed = 19514\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.65,     validation_loss:16.00\n",
      "training_loss:102.64,     validation_loss:16.25\n",
      "training_loss:102.52,     validation_loss:16.20\n",
      "training_loss:102.73,     validation_loss:16.24\n",
      "alpha = 1, hidden layer = 6 seed = 68167\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.77,     validation_loss:16.28\n",
      "training_loss:102.63,     validation_loss:16.20\n",
      "training_loss:102.56,     validation_loss:16.01\n",
      "training_loss:102.66,     validation_loss:16.06\n",
      "alpha = 1, hidden layer = 6 seed = 48275\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.67,     validation_loss:16.38\n",
      "training_loss:102.65,     validation_loss:15.92\n",
      "training_loss:102.73,     validation_loss:16.30\n",
      "alpha = 1, hidden layer = 6 seed = 30650\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.64,     validation_loss:16.34\n",
      "training_loss:102.72,     validation_loss:16.12\n",
      "training_loss:102.56,     validation_loss:15.97\n",
      "training_loss:102.76,     validation_loss:16.06\n",
      "alpha = 1, hidden layer = 6 seed = 54546\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.71,     validation_loss:16.14\n",
      "training_loss:102.53,     validation_loss:16.29\n",
      "training_loss:102.72,     validation_loss:16.20\n",
      "training_loss:102.54,     validation_loss:16.65\n",
      "training_loss:102.54,     validation_loss:16.28\n",
      "training_loss:102.65,     validation_loss:15.93\n",
      "training_loss:102.76,     validation_loss:16.07\n",
      "alpha = 1, hidden layer = 6 seed = 4927\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.65,     validation_loss:16.31\n",
      "training_loss:102.60,     validation_loss:16.49\n",
      "training_loss:102.66,     validation_loss:16.22\n",
      "training_loss:102.64,     validation_loss:16.00\n",
      "training_loss:102.66,     validation_loss:15.97\n",
      "training_loss:102.67,     validation_loss:16.20\n",
      "alpha = 1, hidden layer = 6 seed = 51683\n",
      "Hidden layer size of Refnet is  1819\n",
      "[2, 1819, 1819, 1819, 1819, 1819, 1819, 6129]\n",
      "training_loss:102.65,     validation_loss:16.42\n",
      "training_loss:102.60,     validation_loss:16.24\n",
      "training_loss:102.62,     validation_loss:16.22\n",
      "training_loss:102.62,     validation_loss:16.35\n",
      "alpha = 2.5, hidden layer = 1 seed = 76684\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.71,     validation_loss:16.33\n",
      "training_loss:102.62,     validation_loss:15.94\n",
      "training_loss:102.66,     validation_loss:16.18\n",
      "alpha = 2.5, hidden layer = 1 seed = 86728\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.64,     validation_loss:16.32\n",
      "training_loss:102.63,     validation_loss:16.03\n",
      "training_loss:102.61,     validation_loss:16.64\n",
      "training_loss:102.58,     validation_loss:15.79\n",
      "training_loss:102.68,     validation_loss:16.07\n",
      "alpha = 2.5, hidden layer = 1 seed = 48351\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.55,     validation_loss:16.41\n",
      "training_loss:102.73,     validation_loss:16.37\n",
      "training_loss:102.46,     validation_loss:15.89\n",
      "training_loss:102.68,     validation_loss:16.38\n",
      "alpha = 2.5, hidden layer = 1 seed = 19514\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.58,     validation_loss:16.29\n",
      "training_loss:102.73,     validation_loss:16.35\n",
      "alpha = 2.5, hidden layer = 1 seed = 68167\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.65,     validation_loss:15.78\n",
      "training_loss:102.60,     validation_loss:16.39\n",
      "training_loss:102.51,     validation_loss:16.42\n",
      "training_loss:102.73,     validation_loss:16.32\n",
      "training_loss:102.60,     validation_loss:16.09\n",
      "training_loss:102.55,     validation_loss:16.09\n",
      "training_loss:102.68,     validation_loss:16.51\n",
      "alpha = 2.5, hidden layer = 1 seed = 48275\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.69,     validation_loss:16.44\n",
      "training_loss:102.55,     validation_loss:16.24\n",
      "training_loss:102.65,     validation_loss:16.40\n",
      "alpha = 2.5, hidden layer = 1 seed = 30650\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.64,     validation_loss:16.00\n",
      "training_loss:102.77,     validation_loss:15.75\n",
      "training_loss:102.63,     validation_loss:16.21\n",
      "training_loss:102.71,     validation_loss:16.33\n",
      "alpha = 2.5, hidden layer = 1 seed = 54546\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.73,     validation_loss:16.13\n",
      "training_loss:102.57,     validation_loss:16.30\n",
      "training_loss:102.62,     validation_loss:15.98\n",
      "training_loss:102.75,     validation_loss:16.05\n",
      "alpha = 2.5, hidden layer = 1 seed = 4927\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.68,     validation_loss:16.15\n",
      "training_loss:102.55,     validation_loss:16.07\n",
      "training_loss:102.67,     validation_loss:16.08\n",
      "alpha = 2.5, hidden layer = 1 seed = 51683\n",
      "Hidden layer size of Refnet is  1815\n",
      "[2, 1815, 6129]\n",
      "training_loss:102.62,     validation_loss:16.22\n",
      "training_loss:102.55,     validation_loss:16.62\n",
      "training_loss:102.54,     validation_loss:16.48\n",
      "training_loss:102.60,     validation_loss:16.51\n",
      "alpha = 2.5, hidden layer = 2 seed = 76684\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.62,     validation_loss:16.30\n",
      "training_loss:102.63,     validation_loss:16.14\n",
      "training_loss:102.66,     validation_loss:16.35\n",
      "alpha = 2.5, hidden layer = 2 seed = 86728\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.59,     validation_loss:16.83\n",
      "training_loss:102.61,     validation_loss:16.39\n",
      "training_loss:102.55,     validation_loss:16.14\n",
      "training_loss:102.67,     validation_loss:16.15\n",
      "alpha = 2.5, hidden layer = 2 seed = 48351\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.62,     validation_loss:16.10\n",
      "training_loss:102.71,     validation_loss:16.10\n",
      "alpha = 2.5, hidden layer = 2 seed = 19514\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.66,     validation_loss:16.38\n",
      "training_loss:102.63,     validation_loss:16.28\n",
      "training_loss:102.55,     validation_loss:16.38\n",
      "training_loss:102.57,     validation_loss:16.11\n",
      "training_loss:102.62,     validation_loss:15.97\n",
      "training_loss:102.61,     validation_loss:16.26\n",
      "training_loss:102.69,     validation_loss:16.13\n",
      "training_loss:102.53,     validation_loss:16.18\n",
      "training_loss:102.83,     validation_loss:16.51\n",
      "alpha = 2.5, hidden layer = 2 seed = 68167\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.68,     validation_loss:16.03\n",
      "training_loss:102.64,     validation_loss:16.42\n",
      "training_loss:102.59,     validation_loss:16.65\n",
      "training_loss:102.55,     validation_loss:16.44\n",
      "training_loss:102.54,     validation_loss:16.23\n",
      "training_loss:102.62,     validation_loss:16.02\n",
      "training_loss:102.65,     validation_loss:16.06\n",
      "alpha = 2.5, hidden layer = 2 seed = 48275\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.59,     validation_loss:16.61\n",
      "training_loss:102.64,     validation_loss:16.30\n",
      "training_loss:102.52,     validation_loss:16.44\n",
      "training_loss:102.62,     validation_loss:16.13\n",
      "training_loss:102.64,     validation_loss:16.63\n",
      "alpha = 2.5, hidden layer = 2 seed = 30650\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.55,     validation_loss:16.16\n",
      "training_loss:102.65,     validation_loss:15.96\n",
      "training_loss:102.69,     validation_loss:16.24\n",
      "alpha = 2.5, hidden layer = 2 seed = 54546\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.61,     validation_loss:16.66\n",
      "training_loss:102.66,     validation_loss:15.95\n",
      "training_loss:102.66,     validation_loss:16.23\n",
      "training_loss:102.62,     validation_loss:16.48\n",
      "training_loss:102.59,     validation_loss:16.09\n",
      "training_loss:102.50,     validation_loss:16.39\n",
      "training_loss:102.67,     validation_loss:15.86\n",
      "training_loss:102.65,     validation_loss:16.32\n",
      "training_loss:102.60,     validation_loss:16.26\n",
      "training_loss:102.54,     validation_loss:16.47\n",
      "training_loss:102.63,     validation_loss:16.51\n",
      "alpha = 2.5, hidden layer = 2 seed = 4927\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.56,     validation_loss:16.34\n",
      "training_loss:102.65,     validation_loss:16.31\n",
      "training_loss:102.56,     validation_loss:15.99\n",
      "training_loss:102.69,     validation_loss:16.25\n",
      "alpha = 2.5, hidden layer = 2 seed = 51683\n",
      "Hidden layer size of Refnet is  1845\n",
      "[2, 1845, 1845, 6129]\n",
      "training_loss:102.61,     validation_loss:16.19\n",
      "training_loss:102.71,     validation_loss:15.86\n",
      "training_loss:102.72,     validation_loss:16.13\n",
      "alpha = 2.5, hidden layer = 3 seed = 76684\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.58,     validation_loss:16.57\n",
      "training_loss:102.77,     validation_loss:16.00\n",
      "training_loss:102.57,     validation_loss:16.09\n",
      "training_loss:102.76,     validation_loss:16.38\n",
      "alpha = 2.5, hidden layer = 3 seed = 86728\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.67,     validation_loss:16.23\n",
      "training_loss:102.76,     validation_loss:16.42\n",
      "alpha = 2.5, hidden layer = 3 seed = 48351\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.67,     validation_loss:16.10\n",
      "training_loss:102.70,     validation_loss:16.40\n",
      "alpha = 2.5, hidden layer = 3 seed = 19514\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.68,     validation_loss:16.37\n",
      "training_loss:102.64,     validation_loss:16.32\n",
      "training_loss:102.58,     validation_loss:16.49\n",
      "training_loss:102.61,     validation_loss:16.20\n",
      "training_loss:102.53,     validation_loss:16.00\n",
      "training_loss:102.62,     validation_loss:16.21\n",
      "alpha = 2.5, hidden layer = 3 seed = 68167\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.63,     validation_loss:16.13\n",
      "training_loss:102.62,     validation_loss:16.08\n",
      "training_loss:102.67,     validation_loss:16.21\n",
      "alpha = 2.5, hidden layer = 3 seed = 48275\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.61,     validation_loss:16.19\n",
      "training_loss:102.60,     validation_loss:16.40\n",
      "training_loss:102.57,     validation_loss:16.15\n",
      "training_loss:102.57,     validation_loss:16.50\n",
      "training_loss:102.53,     validation_loss:15.82\n",
      "training_loss:102.62,     validation_loss:16.31\n",
      "alpha = 2.5, hidden layer = 3 seed = 30650\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.66,     validation_loss:16.29\n",
      "training_loss:102.60,     validation_loss:16.34\n",
      "training_loss:102.62,     validation_loss:15.63\n",
      "training_loss:102.79,     validation_loss:16.28\n",
      "alpha = 2.5, hidden layer = 3 seed = 54546\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.72,     validation_loss:15.93\n",
      "training_loss:102.61,     validation_loss:15.88\n",
      "training_loss:102.76,     validation_loss:15.82\n",
      "training_loss:102.60,     validation_loss:15.93\n",
      "training_loss:102.64,     validation_loss:16.78\n",
      "alpha = 2.5, hidden layer = 3 seed = 4927\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.66,     validation_loss:16.02\n",
      "training_loss:102.75,     validation_loss:16.53\n",
      "alpha = 2.5, hidden layer = 3 seed = 51683\n",
      "Hidden layer size of Refnet is  1875\n",
      "[2, 1875, 1875, 1875, 6129]\n",
      "training_loss:102.69,     validation_loss:16.04\n",
      "training_loss:102.57,     validation_loss:16.34\n",
      "training_loss:102.64,     validation_loss:16.75\n",
      "alpha = 2.5, hidden layer = 4 seed = 76684\n",
      "Hidden layer size of Refnet is  1905\n",
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.66,     validation_loss:16.09\n",
      "training_loss:102.53,     validation_loss:16.48\n",
      "training_loss:102.63,     validation_loss:15.87\n",
      "training_loss:102.69,     validation_loss:16.34\n",
      "alpha = 2.5, hidden layer = 4 seed = 86728\n",
      "Hidden layer size of Refnet is  1905\n",
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.71,     validation_loss:15.94\n",
      "training_loss:102.64,     validation_loss:16.03\n",
      "training_loss:102.66,     validation_loss:16.30\n",
      "alpha = 2.5, hidden layer = 4 seed = 48351\n",
      "Hidden layer size of Refnet is  1905\n",
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.67,     validation_loss:16.34\n",
      "training_loss:102.72,     validation_loss:15.88\n",
      "training_loss:102.61,     validation_loss:16.42\n",
      "training_loss:102.69,     validation_loss:15.93\n",
      "training_loss:102.63,     validation_loss:16.16\n",
      "training_loss:102.66,     validation_loss:16.23\n",
      "alpha = 2.5, hidden layer = 4 seed = 19514\n",
      "Hidden layer size of Refnet is  1905\n",
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.68,     validation_loss:15.74\n",
      "training_loss:102.80,     validation_loss:15.94\n",
      "alpha = 2.5, hidden layer = 4 seed = 68167\n",
      "Hidden layer size of Refnet is  1905\n",
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.61,     validation_loss:16.49\n",
      "training_loss:102.63,     validation_loss:16.30\n",
      "training_loss:102.59,     validation_loss:16.23\n",
      "training_loss:102.68,     validation_loss:16.11\n",
      "training_loss:102.67,     validation_loss:15.84\n",
      "training_loss:102.68,     validation_loss:16.32\n",
      "alpha = 2.5, hidden layer = 4 seed = 48275\n",
      "Hidden layer size of Refnet is  1905\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.77,     validation_loss:16.40\n",
      "training_loss:102.53,     validation_loss:16.35\n",
      "training_loss:102.74,     validation_loss:16.14\n",
      "training_loss:102.55,     validation_loss:15.97\n",
      "training_loss:102.77,     validation_loss:16.15\n",
      "alpha = 2.5, hidden layer = 4 seed = 30650\n",
      "Hidden layer size of Refnet is  1905\n",
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.68,     validation_loss:16.42\n",
      "training_loss:102.53,     validation_loss:16.40\n",
      "training_loss:102.84,     validation_loss:16.13\n",
      "training_loss:102.56,     validation_loss:16.46\n",
      "training_loss:102.63,     validation_loss:16.10\n",
      "training_loss:102.61,     validation_loss:16.15\n",
      "training_loss:102.64,     validation_loss:16.00\n",
      "training_loss:102.54,     validation_loss:16.46\n",
      "training_loss:102.65,     validation_loss:16.26\n",
      "training_loss:102.58,     validation_loss:15.96\n",
      "training_loss:102.67,     validation_loss:16.27\n",
      "alpha = 2.5, hidden layer = 4 seed = 54546\n",
      "Hidden layer size of Refnet is  1905\n",
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.72,     validation_loss:16.10\n",
      "training_loss:102.62,     validation_loss:16.36\n",
      "training_loss:102.50,     validation_loss:16.17\n",
      "training_loss:102.72,     validation_loss:16.44\n",
      "alpha = 2.5, hidden layer = 4 seed = 4927\n",
      "Hidden layer size of Refnet is  1905\n",
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.64,     validation_loss:16.48\n",
      "training_loss:102.60,     validation_loss:16.78\n",
      "training_loss:102.54,     validation_loss:16.34\n",
      "training_loss:102.60,     validation_loss:16.63\n",
      "alpha = 2.5, hidden layer = 4 seed = 51683\n",
      "Hidden layer size of Refnet is  1905\n",
      "[2, 1905, 1905, 1905, 1905, 6129]\n",
      "training_loss:102.65,     validation_loss:15.99\n",
      "training_loss:102.72,     validation_loss:16.06\n",
      "alpha = 2.5, hidden layer = 5 seed = 76684\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.71,     validation_loss:16.09\n",
      "training_loss:102.52,     validation_loss:16.23\n",
      "training_loss:102.64,     validation_loss:16.35\n",
      "alpha = 2.5, hidden layer = 5 seed = 86728\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.72,     validation_loss:16.43\n",
      "training_loss:102.59,     validation_loss:16.25\n",
      "training_loss:102.60,     validation_loss:16.15\n",
      "training_loss:102.80,     validation_loss:16.08\n",
      "training_loss:102.52,     validation_loss:16.24\n",
      "training_loss:102.64,     validation_loss:16.48\n",
      "alpha = 2.5, hidden layer = 5 seed = 48351\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.62,     validation_loss:16.39\n",
      "training_loss:102.79,     validation_loss:16.11\n",
      "training_loss:102.62,     validation_loss:16.41\n",
      "training_loss:102.59,     validation_loss:15.81\n",
      "training_loss:102.63,     validation_loss:15.96\n",
      "alpha = 2.5, hidden layer = 5 seed = 19514\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.56,     validation_loss:16.79\n",
      "training_loss:102.67,     validation_loss:16.36\n",
      "training_loss:102.54,     validation_loss:16.29\n",
      "training_loss:102.64,     validation_loss:16.00\n",
      "training_loss:102.60,     validation_loss:15.86\n",
      "training_loss:102.61,     validation_loss:16.37\n",
      "alpha = 2.5, hidden layer = 5 seed = 68167\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.56,     validation_loss:16.23\n",
      "training_loss:102.78,     validation_loss:15.92\n",
      "training_loss:102.51,     validation_loss:16.87\n",
      "training_loss:102.54,     validation_loss:16.10\n",
      "training_loss:102.63,     validation_loss:16.38\n",
      "alpha = 2.5, hidden layer = 5 seed = 48275\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.69,     validation_loss:16.05\n",
      "training_loss:102.65,     validation_loss:15.99\n",
      "training_loss:102.75,     validation_loss:16.23\n",
      "alpha = 2.5, hidden layer = 5 seed = 30650\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.66,     validation_loss:15.90\n",
      "training_loss:102.74,     validation_loss:16.48\n",
      "alpha = 2.5, hidden layer = 5 seed = 54546\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.67,     validation_loss:16.18\n",
      "training_loss:102.66,     validation_loss:16.05\n",
      "training_loss:102.60,     validation_loss:16.12\n",
      "training_loss:102.66,     validation_loss:16.18\n",
      "alpha = 2.5, hidden layer = 5 seed = 4927\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.62,     validation_loss:16.12\n",
      "training_loss:102.72,     validation_loss:16.42\n",
      "alpha = 2.5, hidden layer = 5 seed = 51683\n",
      "Hidden layer size of Refnet is  1935\n",
      "[2, 1935, 1935, 1935, 1935, 1935, 6129]\n",
      "training_loss:102.60,     validation_loss:16.45\n",
      "training_loss:102.62,     validation_loss:16.55\n",
      "alpha = 2.5, hidden layer = 6 seed = 76684\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.61,     validation_loss:16.34\n",
      "training_loss:102.61,     validation_loss:16.24\n",
      "training_loss:102.57,     validation_loss:15.94\n",
      "training_loss:102.58,     validation_loss:16.83\n",
      "alpha = 2.5, hidden layer = 6 seed = 86728\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.64,     validation_loss:16.58\n",
      "training_loss:102.56,     validation_loss:16.69\n",
      "training_loss:102.60,     validation_loss:16.08\n",
      "training_loss:102.62,     validation_loss:16.22\n",
      "alpha = 2.5, hidden layer = 6 seed = 48351\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.63,     validation_loss:16.06\n",
      "training_loss:102.75,     validation_loss:16.15\n",
      "alpha = 2.5, hidden layer = 6 seed = 19514\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.64,     validation_loss:16.49\n",
      "training_loss:102.56,     validation_loss:15.87\n",
      "training_loss:102.67,     validation_loss:16.12\n",
      "alpha = 2.5, hidden layer = 6 seed = 68167\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.66,     validation_loss:16.08\n",
      "training_loss:102.65,     validation_loss:16.65\n",
      "training_loss:102.58,     validation_loss:16.24\n",
      "training_loss:102.57,     validation_loss:16.53\n",
      "training_loss:102.69,     validation_loss:16.26\n",
      "training_loss:102.74,     validation_loss:15.71\n",
      "training_loss:102.60,     validation_loss:15.94\n",
      "training_loss:102.69,     validation_loss:16.44\n",
      "alpha = 2.5, hidden layer = 6 seed = 48275\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.61,     validation_loss:16.33\n",
      "training_loss:102.62,     validation_loss:16.37\n",
      "alpha = 2.5, hidden layer = 6 seed = 30650\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.67,     validation_loss:16.26\n",
      "training_loss:102.68,     validation_loss:15.73\n",
      "training_loss:102.57,     validation_loss:16.61\n",
      "training_loss:102.70,     validation_loss:16.35\n",
      "training_loss:102.49,     validation_loss:16.32\n",
      "training_loss:102.68,     validation_loss:16.17\n",
      "training_loss:102.62,     validation_loss:16.30\n",
      "training_loss:102.56,     validation_loss:16.29\n",
      "training_loss:102.47,     validation_loss:16.40\n",
      "training_loss:102.61,     validation_loss:16.41\n",
      "alpha = 2.5, hidden layer = 6 seed = 54546\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.69,     validation_loss:16.31\n",
      "training_loss:102.58,     validation_loss:16.32\n",
      "training_loss:102.66,     validation_loss:16.34\n",
      "alpha = 2.5, hidden layer = 6 seed = 4927\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.64,     validation_loss:16.24\n",
      "training_loss:102.61,     validation_loss:16.29\n",
      "training_loss:102.52,     validation_loss:15.69\n",
      "training_loss:102.75,     validation_loss:16.76\n",
      "alpha = 2.5, hidden layer = 6 seed = 51683\n",
      "Hidden layer size of Refnet is  1965\n",
      "[2, 1965, 1965, 1965, 1965, 1965, 1965, 6129]\n",
      "training_loss:102.63,     validation_loss:16.09\n",
      "training_loss:102.71,     validation_loss:15.71\n",
      "training_loss:102.70,     validation_loss:16.57\n",
      "training_loss:102.47,     validation_loss:15.99\n",
      "training_loss:102.74,     validation_loss:16.10\n",
      "alpha = 5, hidden layer = 1 seed = 76684\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.71,     validation_loss:16.29\n",
      "training_loss:102.67,     validation_loss:16.53\n",
      "training_loss:102.42,     validation_loss:16.24\n",
      "training_loss:102.66,     validation_loss:16.14\n",
      "training_loss:102.63,     validation_loss:16.47\n",
      "training_loss:102.49,     validation_loss:16.16\n",
      "training_loss:102.69,     validation_loss:16.02\n",
      "training_loss:102.50,     validation_loss:16.10\n",
      "training_loss:102.58,     validation_loss:16.58\n",
      "alpha = 5, hidden layer = 1 seed = 86728\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.68,     validation_loss:15.97\n",
      "training_loss:102.58,     validation_loss:16.38\n",
      "training_loss:102.73,     validation_loss:15.55\n",
      "training_loss:102.70,     validation_loss:16.41\n",
      "training_loss:102.49,     validation_loss:16.42\n",
      "training_loss:102.61,     validation_loss:16.20\n",
      "training_loss:102.66,     validation_loss:16.10\n",
      "training_loss:102.57,     validation_loss:15.95\n",
      "training_loss:102.75,     validation_loss:15.90\n",
      "training_loss:102.65,     validation_loss:16.35\n",
      "training_loss:102.61,     validation_loss:16.47\n",
      "training_loss:102.64,     validation_loss:16.10\n",
      "training_loss:102.61,     validation_loss:16.62\n",
      "training_loss:102.65,     validation_loss:15.92\n",
      "training_loss:102.54,     validation_loss:16.84\n",
      "training_loss:102.65,     validation_loss:15.91\n",
      "training_loss:102.65,     validation_loss:16.15\n",
      "training_loss:102.52,     validation_loss:15.96\n",
      "training_loss:102.81,     validation_loss:16.30\n",
      "alpha = 5, hidden layer = 1 seed = 48351\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.72,     validation_loss:16.33\n",
      "training_loss:102.49,     validation_loss:16.33\n",
      "training_loss:102.62,     validation_loss:15.91\n",
      "training_loss:102.53,     validation_loss:16.24\n",
      "training_loss:102.72,     validation_loss:16.52\n",
      "alpha = 5, hidden layer = 1 seed = 19514\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.71,     validation_loss:15.91\n",
      "training_loss:102.57,     validation_loss:16.44\n",
      "training_loss:102.64,     validation_loss:16.45\n",
      "alpha = 5, hidden layer = 1 seed = 68167\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.63,     validation_loss:16.48\n",
      "training_loss:102.54,     validation_loss:16.44\n",
      "training_loss:102.65,     validation_loss:16.27\n",
      "training_loss:102.63,     validation_loss:16.71\n",
      "training_loss:102.72,     validation_loss:16.19\n",
      "training_loss:102.53,     validation_loss:16.08\n",
      "training_loss:102.84,     validation_loss:16.33\n",
      "alpha = 5, hidden layer = 1 seed = 48275\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.64,     validation_loss:16.56\n",
      "training_loss:102.48,     validation_loss:16.32\n",
      "training_loss:102.58,     validation_loss:16.35\n",
      "alpha = 5, hidden layer = 1 seed = 30650\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.68,     validation_loss:16.15\n",
      "training_loss:102.63,     validation_loss:16.17\n",
      "training_loss:102.75,     validation_loss:16.03\n",
      "training_loss:102.63,     validation_loss:16.07\n",
      "training_loss:102.70,     validation_loss:16.28\n",
      "alpha = 5, hidden layer = 1 seed = 54546\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.70,     validation_loss:16.09\n",
      "training_loss:102.57,     validation_loss:16.37\n",
      "training_loss:102.67,     validation_loss:15.58\n",
      "training_loss:102.70,     validation_loss:16.62\n",
      "alpha = 5, hidden layer = 1 seed = 4927\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.61,     validation_loss:16.35\n",
      "training_loss:102.73,     validation_loss:15.96\n",
      "training_loss:102.64,     validation_loss:16.24\n",
      "training_loss:102.53,     validation_loss:16.39\n",
      "training_loss:102.53,     validation_loss:16.44\n",
      "alpha = 5, hidden layer = 1 seed = 51683\n",
      "Hidden layer size of Refnet is  1896\n",
      "[2, 1896, 6129]\n",
      "training_loss:102.67,     validation_loss:15.99\n",
      "training_loss:102.76,     validation_loss:16.09\n",
      "alpha = 5, hidden layer = 2 seed = 76684\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.66,     validation_loss:15.94\n",
      "training_loss:102.67,     validation_loss:16.12\n",
      "alpha = 5, hidden layer = 2 seed = 86728\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.60,     validation_loss:16.46\n",
      "training_loss:102.58,     validation_loss:15.95\n",
      "training_loss:102.74,     validation_loss:15.98\n",
      "alpha = 5, hidden layer = 2 seed = 48351\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.61,     validation_loss:16.39\n",
      "training_loss:102.68,     validation_loss:16.07\n",
      "training_loss:102.53,     validation_loss:16.12\n",
      "training_loss:102.65,     validation_loss:16.38\n",
      "alpha = 5, hidden layer = 2 seed = 19514\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.61,     validation_loss:16.65\n",
      "training_loss:102.56,     validation_loss:16.47\n",
      "training_loss:102.53,     validation_loss:16.30\n",
      "training_loss:102.70,     validation_loss:15.96\n",
      "training_loss:102.58,     validation_loss:16.20\n",
      "training_loss:102.56,     validation_loss:16.07\n",
      "training_loss:102.65,     validation_loss:16.22\n",
      "alpha = 5, hidden layer = 2 seed = 68167\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.63,     validation_loss:16.37\n",
      "training_loss:102.49,     validation_loss:16.53\n",
      "training_loss:102.71,     validation_loss:16.35\n",
      "training_loss:102.61,     validation_loss:16.18\n",
      "training_loss:102.55,     validation_loss:16.51\n",
      "training_loss:102.61,     validation_loss:16.05\n",
      "training_loss:102.80,     validation_loss:16.13\n",
      "alpha = 5, hidden layer = 2 seed = 48275\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.67,     validation_loss:16.14\n",
      "training_loss:102.55,     validation_loss:16.29\n",
      "training_loss:102.69,     validation_loss:16.33\n",
      "alpha = 5, hidden layer = 2 seed = 30650\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.68,     validation_loss:15.95\n",
      "training_loss:102.75,     validation_loss:16.39\n",
      "alpha = 5, hidden layer = 2 seed = 54546\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.70,     validation_loss:16.29\n",
      "training_loss:102.57,     validation_loss:16.43\n",
      "training_loss:102.58,     validation_loss:16.25\n",
      "training_loss:102.68,     validation_loss:16.47\n",
      "alpha = 5, hidden layer = 2 seed = 4927\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.65,     validation_loss:16.02\n",
      "training_loss:102.65,     validation_loss:16.01\n",
      "training_loss:102.60,     validation_loss:16.04\n",
      "training_loss:102.51,     validation_loss:15.97\n",
      "training_loss:102.59,     validation_loss:16.87\n",
      "alpha = 5, hidden layer = 2 seed = 51683\n",
      "Hidden layer size of Refnet is  2005\n",
      "[2, 2005, 2005, 6129]\n",
      "training_loss:102.61,     validation_loss:16.23\n",
      "training_loss:102.71,     validation_loss:16.26\n",
      "alpha = 5, hidden layer = 3 seed = 76684\n",
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.59,     validation_loss:16.73\n",
      "training_loss:102.69,     validation_loss:16.48\n",
      "training_loss:102.47,     validation_loss:16.57\n",
      "training_loss:102.68,     validation_loss:16.44\n",
      "training_loss:102.48,     validation_loss:16.47\n",
      "training_loss:102.56,     validation_loss:16.10\n",
      "training_loss:102.67,     validation_loss:16.23\n",
      "alpha = 5, hidden layer = 3 seed = 86728\n",
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.71,     validation_loss:15.95\n",
      "training_loss:102.69,     validation_loss:15.87\n",
      "training_loss:102.72,     validation_loss:15.95\n",
      "alpha = 5, hidden layer = 3 seed = 48351\n",
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.62,     validation_loss:16.43\n",
      "training_loss:102.62,     validation_loss:16.48\n",
      "training_loss:102.63,     validation_loss:15.87\n",
      "training_loss:102.73,     validation_loss:16.25\n",
      "alpha = 5, hidden layer = 3 seed = 19514\n",
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.64,     validation_loss:16.12\n",
      "training_loss:102.60,     validation_loss:16.44\n",
      "training_loss:102.64,     validation_loss:16.31\n",
      "training_loss:102.60,     validation_loss:15.86\n",
      "training_loss:102.71,     validation_loss:16.71\n",
      "alpha = 5, hidden layer = 3 seed = 68167\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.68,     validation_loss:16.61\n",
      "training_loss:102.43,     validation_loss:16.23\n",
      "training_loss:102.69,     validation_loss:16.20\n",
      "training_loss:102.64,     validation_loss:15.41\n",
      "training_loss:102.76,     validation_loss:16.49\n",
      "alpha = 5, hidden layer = 3 seed = 48275\n",
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.62,     validation_loss:16.09\n",
      "training_loss:102.67,     validation_loss:16.52\n",
      "alpha = 5, hidden layer = 3 seed = 30650\n",
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.64,     validation_loss:16.03\n",
      "training_loss:102.54,     validation_loss:16.39\n",
      "training_loss:102.67,     validation_loss:16.25\n",
      "training_loss:102.54,     validation_loss:16.43\n",
      "training_loss:102.69,     validation_loss:16.05\n",
      "training_loss:102.61,     validation_loss:16.30\n",
      "training_loss:102.55,     validation_loss:16.05\n",
      "training_loss:102.68,     validation_loss:16.12\n",
      "alpha = 5, hidden layer = 3 seed = 54546\n",
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.71,     validation_loss:16.28\n",
      "training_loss:102.61,     validation_loss:16.19\n",
      "training_loss:102.67,     validation_loss:16.33\n",
      "alpha = 5, hidden layer = 3 seed = 4927\n",
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.67,     validation_loss:16.64\n",
      "training_loss:102.57,     validation_loss:16.15\n",
      "training_loss:102.65,     validation_loss:16.11\n",
      "training_loss:102.57,     validation_loss:16.34\n",
      "training_loss:102.56,     validation_loss:16.19\n",
      "training_loss:102.70,     validation_loss:16.08\n",
      "training_loss:102.57,     validation_loss:16.15\n",
      "training_loss:102.70,     validation_loss:16.21\n",
      "alpha = 5, hidden layer = 3 seed = 51683\n",
      "Hidden layer size of Refnet is  2113\n",
      "[2, 2113, 2113, 2113, 6129]\n",
      "training_loss:102.67,     validation_loss:16.40\n",
      "training_loss:102.72,     validation_loss:15.78\n",
      "training_loss:102.65,     validation_loss:16.57\n",
      "training_loss:102.63,     validation_loss:16.13\n",
      "training_loss:102.58,     validation_loss:16.33\n",
      "training_loss:102.61,     validation_loss:16.16\n",
      "training_loss:102.60,     validation_loss:16.17\n",
      "training_loss:102.62,     validation_loss:16.09\n",
      "training_loss:102.62,     validation_loss:16.65\n",
      "training_loss:102.56,     validation_loss:16.32\n",
      "training_loss:102.59,     validation_loss:16.22\n",
      "training_loss:102.61,     validation_loss:15.94\n",
      "training_loss:102.78,     validation_loss:16.20\n",
      "alpha = 5, hidden layer = 4 seed = 76684\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.69,     validation_loss:16.24\n",
      "training_loss:102.50,     validation_loss:16.40\n",
      "training_loss:102.63,     validation_loss:16.13\n",
      "training_loss:102.55,     validation_loss:16.20\n",
      "training_loss:102.63,     validation_loss:16.06\n",
      "training_loss:102.59,     validation_loss:15.95\n",
      "training_loss:102.65,     validation_loss:16.37\n",
      "alpha = 5, hidden layer = 4 seed = 86728\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.71,     validation_loss:16.31\n",
      "training_loss:102.72,     validation_loss:16.15\n",
      "training_loss:102.53,     validation_loss:16.32\n",
      "training_loss:102.63,     validation_loss:15.68\n",
      "training_loss:102.64,     validation_loss:16.15\n",
      "alpha = 5, hidden layer = 4 seed = 48351\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.65,     validation_loss:16.13\n",
      "training_loss:102.70,     validation_loss:16.28\n",
      "alpha = 5, hidden layer = 4 seed = 19514\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.61,     validation_loss:16.70\n",
      "training_loss:102.50,     validation_loss:16.16\n",
      "training_loss:102.54,     validation_loss:16.42\n",
      "alpha = 5, hidden layer = 4 seed = 68167\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.63,     validation_loss:16.08\n",
      "training_loss:102.73,     validation_loss:16.41\n",
      "alpha = 5, hidden layer = 4 seed = 48275\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.72,     validation_loss:16.11\n",
      "training_loss:102.56,     validation_loss:16.45\n",
      "training_loss:102.55,     validation_loss:16.17\n",
      "training_loss:102.70,     validation_loss:16.22\n",
      "alpha = 5, hidden layer = 4 seed = 30650\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.71,     validation_loss:16.02\n",
      "training_loss:102.67,     validation_loss:16.31\n",
      "training_loss:102.45,     validation_loss:16.18\n",
      "training_loss:102.78,     validation_loss:15.87\n",
      "training_loss:102.65,     validation_loss:15.73\n",
      "training_loss:102.69,     validation_loss:16.10\n",
      "alpha = 5, hidden layer = 4 seed = 54546\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.73,     validation_loss:16.05\n",
      "training_loss:102.58,     validation_loss:15.81\n",
      "training_loss:102.81,     validation_loss:16.13\n",
      "alpha = 5, hidden layer = 4 seed = 4927\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.65,     validation_loss:16.14\n",
      "training_loss:102.67,     validation_loss:16.29\n",
      "alpha = 5, hidden layer = 4 seed = 51683\n",
      "Hidden layer size of Refnet is  2220\n",
      "[2, 2220, 2220, 2220, 2220, 6129]\n",
      "training_loss:102.76,     validation_loss:16.12\n",
      "training_loss:102.59,     validation_loss:16.26\n",
      "training_loss:102.65,     validation_loss:15.92\n",
      "training_loss:102.73,     validation_loss:16.03\n",
      "alpha = 5, hidden layer = 5 seed = 76684\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.70,     validation_loss:16.23\n",
      "training_loss:102.62,     validation_loss:16.50\n",
      "training_loss:102.58,     validation_loss:15.74\n",
      "training_loss:102.66,     validation_loss:16.39\n",
      "alpha = 5, hidden layer = 5 seed = 86728\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.62,     validation_loss:16.15\n",
      "training_loss:102.73,     validation_loss:16.48\n",
      "alpha = 5, hidden layer = 5 seed = 48351\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.73,     validation_loss:15.82\n",
      "training_loss:102.61,     validation_loss:16.48\n",
      "training_loss:102.67,     validation_loss:16.25\n",
      "training_loss:102.65,     validation_loss:15.86\n",
      "training_loss:102.63,     validation_loss:16.50\n",
      "training_loss:102.71,     validation_loss:15.94\n",
      "training_loss:102.63,     validation_loss:16.47\n",
      "training_loss:102.55,     validation_loss:15.66\n",
      "training_loss:102.67,     validation_loss:16.34\n",
      "alpha = 5, hidden layer = 5 seed = 19514\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.68,     validation_loss:16.32\n",
      "training_loss:102.51,     validation_loss:16.33\n",
      "training_loss:102.60,     validation_loss:16.01\n",
      "training_loss:102.58,     validation_loss:16.13\n",
      "training_loss:102.70,     validation_loss:16.04\n",
      "training_loss:102.63,     validation_loss:16.26\n",
      "training_loss:102.60,     validation_loss:16.29\n",
      "training_loss:102.66,     validation_loss:16.47\n",
      "alpha = 5, hidden layer = 5 seed = 68167\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.61,     validation_loss:16.59\n",
      "training_loss:102.56,     validation_loss:15.95\n",
      "training_loss:102.70,     validation_loss:16.11\n",
      "alpha = 5, hidden layer = 5 seed = 48275\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.63,     validation_loss:16.18\n",
      "training_loss:102.62,     validation_loss:16.38\n",
      "training_loss:102.57,     validation_loss:16.21\n",
      "training_loss:102.54,     validation_loss:15.87\n",
      "training_loss:102.68,     validation_loss:15.92\n",
      "alpha = 5, hidden layer = 5 seed = 30650\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.60,     validation_loss:16.26\n",
      "training_loss:102.64,     validation_loss:16.03\n",
      "training_loss:102.62,     validation_loss:16.47\n",
      "training_loss:102.62,     validation_loss:16.42\n",
      "training_loss:102.51,     validation_loss:15.94\n",
      "training_loss:102.75,     validation_loss:16.44\n",
      "alpha = 5, hidden layer = 5 seed = 54546\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.66,     validation_loss:16.26\n",
      "training_loss:102.63,     validation_loss:16.51\n",
      "training_loss:102.50,     validation_loss:16.16\n",
      "training_loss:102.64,     validation_loss:16.43\n",
      "alpha = 5, hidden layer = 5 seed = 4927\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.59,     validation_loss:16.02\n",
      "training_loss:102.68,     validation_loss:16.15\n",
      "alpha = 5, hidden layer = 5 seed = 51683\n",
      "Hidden layer size of Refnet is  2325\n",
      "[2, 2325, 2325, 2325, 2325, 2325, 6129]\n",
      "training_loss:102.63,     validation_loss:16.15\n",
      "training_loss:102.59,     validation_loss:16.35\n",
      "training_loss:102.62,     validation_loss:16.46\n",
      "alpha = 5, hidden layer = 6 seed = 76684\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.58,     validation_loss:16.48\n",
      "training_loss:102.65,     validation_loss:16.50\n",
      "alpha = 5, hidden layer = 6 seed = 86728\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.66,     validation_loss:16.13\n",
      "training_loss:102.42,     validation_loss:16.00\n",
      "training_loss:102.66,     validation_loss:16.30\n",
      "alpha = 5, hidden layer = 6 seed = 48351\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.70,     validation_loss:16.26\n",
      "training_loss:102.70,     validation_loss:16.24\n",
      "training_loss:102.69,     validation_loss:16.03\n",
      "training_loss:102.70,     validation_loss:16.20\n",
      "alpha = 5, hidden layer = 6 seed = 19514\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.66,     validation_loss:16.17\n",
      "training_loss:102.76,     validation_loss:16.34\n",
      "alpha = 5, hidden layer = 6 seed = 68167\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.73,     validation_loss:16.33\n",
      "training_loss:102.53,     validation_loss:16.02\n",
      "training_loss:102.66,     validation_loss:16.67\n",
      "alpha = 5, hidden layer = 6 seed = 48275\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.65,     validation_loss:16.04\n",
      "training_loss:102.75,     validation_loss:16.22\n",
      "alpha = 5, hidden layer = 6 seed = 30650\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.73,     validation_loss:16.01\n",
      "training_loss:102.75,     validation_loss:16.17\n",
      "alpha = 5, hidden layer = 6 seed = 54546\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.67,     validation_loss:16.45\n",
      "training_loss:102.52,     validation_loss:16.51\n",
      "training_loss:102.68,     validation_loss:16.05\n",
      "training_loss:102.60,     validation_loss:16.30\n",
      "training_loss:102.73,     validation_loss:16.22\n",
      "training_loss:102.65,     validation_loss:16.57\n",
      "training_loss:102.63,     validation_loss:16.03\n",
      "training_loss:102.66,     validation_loss:16.28\n",
      "alpha = 5, hidden layer = 6 seed = 4927\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.76,     validation_loss:15.94\n",
      "training_loss:102.67,     validation_loss:16.32\n",
      "training_loss:102.64,     validation_loss:16.21\n",
      "training_loss:102.54,     validation_loss:16.49\n",
      "training_loss:102.73,     validation_loss:15.78\n",
      "training_loss:102.64,     validation_loss:16.32\n",
      "training_loss:102.74,     validation_loss:15.98\n",
      "training_loss:102.56,     validation_loss:16.33\n",
      "training_loss:102.65,     validation_loss:16.09\n",
      "training_loss:102.61,     validation_loss:16.46\n",
      "training_loss:102.53,     validation_loss:16.25\n",
      "training_loss:102.67,     validation_loss:16.13\n",
      "training_loss:102.70,     validation_loss:16.61\n",
      "alpha = 5, hidden layer = 6 seed = 51683\n",
      "Hidden layer size of Refnet is  2429\n",
      "[2, 2429, 2429, 2429, 2429, 2429, 2429, 6129]\n",
      "training_loss:102.68,     validation_loss:16.30\n",
      "training_loss:102.62,     validation_loss:16.08\n",
      "training_loss:102.74,     validation_loss:16.37\n",
      "alpha = 10, hidden layer = 1 seed = 76684\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.69,     validation_loss:16.01\n",
      "training_loss:102.63,     validation_loss:16.61\n",
      "training_loss:102.57,     validation_loss:15.61\n",
      "training_loss:102.71,     validation_loss:16.48\n",
      "alpha = 10, hidden layer = 1 seed = 86728\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.64,     validation_loss:15.93\n",
      "training_loss:102.73,     validation_loss:16.48\n",
      "alpha = 10, hidden layer = 1 seed = 48351\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.64,     validation_loss:16.59\n",
      "training_loss:102.67,     validation_loss:15.73\n",
      "training_loss:102.78,     validation_loss:16.20\n",
      "alpha = 10, hidden layer = 1 seed = 19514\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.59,     validation_loss:16.29\n",
      "training_loss:102.74,     validation_loss:15.80\n",
      "training_loss:102.63,     validation_loss:16.10\n",
      "training_loss:102.59,     validation_loss:16.36\n",
      "training_loss:102.51,     validation_loss:16.04\n",
      "training_loss:102.67,     validation_loss:16.14\n",
      "alpha = 10, hidden layer = 1 seed = 68167\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.64,     validation_loss:16.35\n",
      "training_loss:102.64,     validation_loss:16.27\n",
      "training_loss:102.52,     validation_loss:16.37\n",
      "training_loss:102.69,     validation_loss:15.76\n",
      "training_loss:102.60,     validation_loss:16.27\n",
      "training_loss:102.63,     validation_loss:16.64\n",
      "alpha = 10, hidden layer = 1 seed = 48275\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.63,     validation_loss:16.62\n",
      "training_loss:102.58,     validation_loss:16.33\n",
      "training_loss:102.55,     validation_loss:16.46\n",
      "training_loss:102.58,     validation_loss:16.39\n",
      "training_loss:102.67,     validation_loss:16.23\n",
      "training_loss:102.60,     validation_loss:16.09\n",
      "training_loss:102.66,     validation_loss:16.20\n",
      "alpha = 10, hidden layer = 1 seed = 30650\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.68,     validation_loss:16.18\n",
      "training_loss:102.65,     validation_loss:16.21\n",
      "training_loss:102.63,     validation_loss:16.31\n",
      "training_loss:102.58,     validation_loss:16.26\n",
      "training_loss:102.58,     validation_loss:16.26\n",
      "training_loss:102.53,     validation_loss:16.24\n",
      "training_loss:102.74,     validation_loss:16.33\n",
      "alpha = 10, hidden layer = 1 seed = 54546\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.71,     validation_loss:16.15\n",
      "training_loss:102.61,     validation_loss:16.17\n",
      "training_loss:102.67,     validation_loss:16.34\n",
      "alpha = 10, hidden layer = 1 seed = 4927\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.70,     validation_loss:16.40\n",
      "training_loss:102.64,     validation_loss:16.15\n",
      "training_loss:102.58,     validation_loss:16.09\n",
      "training_loss:102.70,     validation_loss:16.10\n",
      "alpha = 10, hidden layer = 1 seed = 51683\n",
      "Hidden layer size of Refnet is  2149\n",
      "[2, 2149, 6129]\n",
      "training_loss:102.69,     validation_loss:16.01\n",
      "training_loss:102.57,     validation_loss:16.29\n",
      "training_loss:102.71,     validation_loss:15.90\n",
      "training_loss:102.65,     validation_loss:16.37\n",
      "training_loss:102.56,     validation_loss:16.21\n",
      "training_loss:102.67,     validation_loss:16.17\n",
      "training_loss:102.65,     validation_loss:16.11\n",
      "training_loss:102.63,     validation_loss:16.11\n",
      "training_loss:102.67,     validation_loss:15.72\n",
      "training_loss:102.71,     validation_loss:16.10\n",
      "alpha = 10, hidden layer = 2 seed = 76684\n",
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.66,     validation_loss:16.73\n",
      "training_loss:102.58,     validation_loss:16.20\n",
      "training_loss:102.72,     validation_loss:15.96\n",
      "training_loss:102.58,     validation_loss:16.35\n",
      "training_loss:102.63,     validation_loss:16.31\n",
      "training_loss:102.60,     validation_loss:16.01\n",
      "training_loss:102.73,     validation_loss:15.94\n",
      "training_loss:102.68,     validation_loss:16.33\n",
      "training_loss:102.52,     validation_loss:15.80\n",
      "training_loss:102.70,     validation_loss:16.36\n",
      "alpha = 10, hidden layer = 2 seed = 86728\n",
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.69,     validation_loss:16.08\n",
      "training_loss:102.61,     validation_loss:15.57\n",
      "training_loss:102.79,     validation_loss:16.22\n",
      "alpha = 10, hidden layer = 2 seed = 48351\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.67,     validation_loss:16.19\n",
      "training_loss:102.59,     validation_loss:16.14\n",
      "training_loss:102.61,     validation_loss:16.43\n",
      "alpha = 10, hidden layer = 2 seed = 19514\n",
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.56,     validation_loss:16.36\n",
      "training_loss:102.72,     validation_loss:16.28\n",
      "training_loss:102.36,     validation_loss:16.17\n",
      "training_loss:102.84,     validation_loss:16.39\n",
      "alpha = 10, hidden layer = 2 seed = 68167\n",
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.62,     validation_loss:16.34\n",
      "training_loss:102.51,     validation_loss:16.18\n",
      "training_loss:102.72,     validation_loss:16.53\n",
      "alpha = 10, hidden layer = 2 seed = 48275\n",
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.64,     validation_loss:16.49\n",
      "training_loss:102.56,     validation_loss:16.50\n",
      "training_loss:102.73,     validation_loss:16.05\n",
      "training_loss:102.62,     validation_loss:15.96\n",
      "training_loss:102.73,     validation_loss:16.32\n",
      "alpha = 10, hidden layer = 2 seed = 30650\n",
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.69,     validation_loss:16.40\n",
      "training_loss:102.60,     validation_loss:16.34\n",
      "training_loss:102.62,     validation_loss:16.13\n",
      "training_loss:102.68,     validation_loss:16.20\n",
      "alpha = 10, hidden layer = 2 seed = 54546\n",
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.66,     validation_loss:15.76\n",
      "training_loss:102.74,     validation_loss:16.29\n",
      "alpha = 10, hidden layer = 2 seed = 4927\n",
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.52,     validation_loss:15.89\n",
      "training_loss:102.73,     validation_loss:16.02\n",
      "alpha = 10, hidden layer = 2 seed = 51683\n",
      "Hidden layer size of Refnet is  2498\n",
      "[2, 2498, 2498, 6129]\n",
      "training_loss:102.62,     validation_loss:16.25\n",
      "training_loss:102.64,     validation_loss:16.01\n",
      "training_loss:102.60,     validation_loss:16.19\n",
      "training_loss:102.62,     validation_loss:16.41\n",
      "alpha = 10, hidden layer = 3 seed = 76684\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.70,     validation_loss:16.05\n",
      "training_loss:102.68,     validation_loss:16.75\n",
      "training_loss:102.55,     validation_loss:16.00\n",
      "training_loss:102.66,     validation_loss:16.49\n",
      "alpha = 10, hidden layer = 3 seed = 86728\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.64,     validation_loss:16.47\n",
      "training_loss:102.46,     validation_loss:16.33\n",
      "training_loss:102.71,     validation_loss:16.22\n",
      "training_loss:102.62,     validation_loss:16.47\n",
      "training_loss:102.49,     validation_loss:16.11\n",
      "training_loss:102.63,     validation_loss:16.04\n",
      "training_loss:102.60,     validation_loss:15.99\n",
      "training_loss:102.66,     validation_loss:15.98\n",
      "training_loss:102.59,     validation_loss:16.03\n",
      "training_loss:102.63,     validation_loss:16.37\n",
      "alpha = 10, hidden layer = 3 seed = 48351\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.68,     validation_loss:16.02\n",
      "training_loss:102.66,     validation_loss:16.14\n",
      "training_loss:102.63,     validation_loss:16.20\n",
      "training_loss:102.70,     validation_loss:16.30\n",
      "alpha = 10, hidden layer = 3 seed = 19514\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.55,     validation_loss:16.34\n",
      "training_loss:102.70,     validation_loss:16.22\n",
      "training_loss:102.65,     validation_loss:16.27\n",
      "training_loss:102.65,     validation_loss:16.36\n",
      "training_loss:102.62,     validation_loss:16.15\n",
      "training_loss:102.54,     validation_loss:16.52\n",
      "training_loss:102.56,     validation_loss:16.29\n",
      "training_loss:102.60,     validation_loss:16.39\n",
      "alpha = 10, hidden layer = 3 seed = 68167\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.70,     validation_loss:16.31\n",
      "training_loss:102.58,     validation_loss:16.51\n",
      "training_loss:102.63,     validation_loss:16.23\n",
      "training_loss:102.72,     validation_loss:16.24\n",
      "alpha = 10, hidden layer = 3 seed = 48275\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.75,     validation_loss:16.06\n",
      "training_loss:102.59,     validation_loss:16.21\n",
      "training_loss:102.69,     validation_loss:15.93\n",
      "training_loss:102.73,     validation_loss:15.90\n",
      "training_loss:102.58,     validation_loss:15.97\n",
      "training_loss:102.65,     validation_loss:16.27\n",
      "alpha = 10, hidden layer = 3 seed = 30650\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.68,     validation_loss:16.31\n",
      "training_loss:102.67,     validation_loss:16.19\n",
      "training_loss:102.63,     validation_loss:15.98\n",
      "training_loss:102.59,     validation_loss:16.28\n",
      "training_loss:102.78,     validation_loss:16.09\n",
      "training_loss:102.51,     validation_loss:15.77\n",
      "training_loss:102.77,     validation_loss:16.39\n",
      "alpha = 10, hidden layer = 3 seed = 54546\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.59,     validation_loss:16.14\n",
      "training_loss:102.76,     validation_loss:15.99\n",
      "training_loss:102.59,     validation_loss:15.93\n",
      "training_loss:102.74,     validation_loss:16.64\n",
      "alpha = 10, hidden layer = 3 seed = 4927\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.64,     validation_loss:16.02\n",
      "training_loss:102.61,     validation_loss:15.92\n",
      "training_loss:102.69,     validation_loss:16.48\n",
      "alpha = 10, hidden layer = 3 seed = 51683\n",
      "Hidden layer size of Refnet is  2834\n",
      "[2, 2834, 2834, 2834, 6129]\n",
      "training_loss:102.60,     validation_loss:16.19\n",
      "training_loss:102.69,     validation_loss:16.16\n",
      "training_loss:102.57,     validation_loss:15.99\n",
      "training_loss:102.79,     validation_loss:16.39\n",
      "alpha = 10, hidden layer = 4 seed = 76684\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.69,     validation_loss:16.11\n",
      "training_loss:102.48,     validation_loss:16.52\n",
      "training_loss:102.59,     validation_loss:16.34\n",
      "training_loss:102.65,     validation_loss:16.26\n",
      "training_loss:102.53,     validation_loss:16.11\n",
      "training_loss:102.73,     validation_loss:16.04\n",
      "training_loss:102.46,     validation_loss:16.53\n",
      "training_loss:102.62,     validation_loss:16.72\n",
      "alpha = 10, hidden layer = 4 seed = 86728\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.69,     validation_loss:16.05\n",
      "training_loss:102.53,     validation_loss:16.18\n",
      "training_loss:102.73,     validation_loss:16.16\n",
      "training_loss:102.56,     validation_loss:16.22\n",
      "training_loss:102.55,     validation_loss:16.34\n",
      "training_loss:102.64,     validation_loss:15.91\n",
      "training_loss:102.61,     validation_loss:16.10\n",
      "training_loss:102.64,     validation_loss:16.40\n",
      "alpha = 10, hidden layer = 4 seed = 48351\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.67,     validation_loss:16.04\n",
      "training_loss:102.62,     validation_loss:16.31\n",
      "training_loss:102.64,     validation_loss:16.03\n",
      "training_loss:102.57,     validation_loss:16.76\n",
      "training_loss:102.61,     validation_loss:16.21\n",
      "training_loss:102.48,     validation_loss:16.28\n",
      "training_loss:102.60,     validation_loss:16.39\n",
      "alpha = 10, hidden layer = 4 seed = 19514\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.65,     validation_loss:16.06\n",
      "training_loss:102.77,     validation_loss:16.12\n",
      "alpha = 10, hidden layer = 4 seed = 68167\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.57,     validation_loss:16.10\n",
      "training_loss:102.64,     validation_loss:16.57\n",
      "alpha = 10, hidden layer = 4 seed = 48275\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.57,     validation_loss:16.00\n",
      "training_loss:102.71,     validation_loss:16.12\n",
      "alpha = 10, hidden layer = 4 seed = 30650\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.65,     validation_loss:16.35\n",
      "training_loss:102.67,     validation_loss:16.25\n",
      "training_loss:102.54,     validation_loss:16.52\n",
      "training_loss:102.63,     validation_loss:16.56\n",
      "alpha = 10, hidden layer = 4 seed = 54546\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.66,     validation_loss:16.30\n",
      "training_loss:102.63,     validation_loss:16.35\n",
      "training_loss:102.60,     validation_loss:16.03\n",
      "training_loss:102.55,     validation_loss:16.78\n",
      "training_loss:102.63,     validation_loss:16.15\n",
      "training_loss:102.53,     validation_loss:16.03\n",
      "training_loss:102.71,     validation_loss:16.59\n",
      "alpha = 10, hidden layer = 4 seed = 4927\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.63,     validation_loss:16.62\n",
      "training_loss:102.56,     validation_loss:16.96\n",
      "training_loss:102.54,     validation_loss:15.90\n",
      "training_loss:102.70,     validation_loss:16.17\n",
      "alpha = 10, hidden layer = 4 seed = 51683\n",
      "Hidden layer size of Refnet is  3157\n",
      "[2, 3157, 3157, 3157, 3157, 6129]\n",
      "training_loss:102.63,     validation_loss:16.03\n",
      "training_loss:102.63,     validation_loss:16.29\n",
      "alpha = 10, hidden layer = 5 seed = 76684\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.64,     validation_loss:16.43\n",
      "training_loss:102.69,     validation_loss:16.44\n",
      "alpha = 10, hidden layer = 5 seed = 86728\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.73,     validation_loss:15.81\n",
      "training_loss:102.88,     validation_loss:16.12\n",
      "alpha = 10, hidden layer = 5 seed = 48351\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.64,     validation_loss:16.44\n",
      "training_loss:102.72,     validation_loss:16.35\n",
      "training_loss:102.46,     validation_loss:16.47\n",
      "training_loss:102.67,     validation_loss:15.95\n",
      "training_loss:102.64,     validation_loss:15.96\n",
      "training_loss:102.59,     validation_loss:15.76\n",
      "training_loss:102.69,     validation_loss:16.15\n",
      "alpha = 10, hidden layer = 5 seed = 19514\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.63,     validation_loss:16.63\n",
      "training_loss:102.59,     validation_loss:16.51\n",
      "training_loss:102.60,     validation_loss:15.75\n",
      "training_loss:102.67,     validation_loss:16.89\n",
      "alpha = 10, hidden layer = 5 seed = 68167\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.71,     validation_loss:16.10\n",
      "training_loss:102.55,     validation_loss:16.45\n",
      "training_loss:102.64,     validation_loss:16.24\n",
      "training_loss:102.66,     validation_loss:15.66\n",
      "training_loss:102.78,     validation_loss:15.86\n",
      "alpha = 10, hidden layer = 5 seed = 48275\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.71,     validation_loss:15.99\n",
      "training_loss:102.68,     validation_loss:16.20\n",
      "training_loss:102.65,     validation_loss:16.17\n",
      "training_loss:102.57,     validation_loss:16.21\n",
      "training_loss:102.67,     validation_loss:16.36\n",
      "alpha = 10, hidden layer = 5 seed = 30650\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.67,     validation_loss:16.22\n",
      "training_loss:102.70,     validation_loss:16.29\n",
      "alpha = 10, hidden layer = 5 seed = 54546\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.71,     validation_loss:16.17\n",
      "training_loss:102.60,     validation_loss:15.98\n",
      "training_loss:102.70,     validation_loss:16.25\n",
      "alpha = 10, hidden layer = 5 seed = 4927\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.68,     validation_loss:16.35\n",
      "training_loss:102.66,     validation_loss:16.31\n",
      "training_loss:102.65,     validation_loss:16.24\n",
      "training_loss:102.66,     validation_loss:15.99\n",
      "training_loss:102.58,     validation_loss:16.27\n",
      "training_loss:102.69,     validation_loss:15.99\n",
      "training_loss:102.59,     validation_loss:16.09\n",
      "training_loss:102.74,     validation_loss:16.04\n",
      "training_loss:102.62,     validation_loss:16.03\n",
      "training_loss:102.69,     validation_loss:15.90\n",
      "training_loss:102.67,     validation_loss:16.23\n",
      "training_loss:102.59,     validation_loss:15.90\n",
      "training_loss:102.72,     validation_loss:16.43\n",
      "alpha = 10, hidden layer = 5 seed = 51683\n",
      "Hidden layer size of Refnet is  3469\n",
      "[2, 3469, 3469, 3469, 3469, 3469, 6129]\n",
      "training_loss:102.69,     validation_loss:16.17\n",
      "training_loss:102.68,     validation_loss:16.17\n",
      "training_loss:102.60,     validation_loss:16.21\n",
      "training_loss:102.74,     validation_loss:15.86\n",
      "training_loss:102.60,     validation_loss:16.28\n",
      "training_loss:102.61,     validation_loss:16.37\n",
      "alpha = 10, hidden layer = 6 seed = 76684\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.58,     validation_loss:15.94\n",
      "training_loss:102.86,     validation_loss:16.29\n",
      "alpha = 10, hidden layer = 6 seed = 86728\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.66,     validation_loss:16.13\n",
      "training_loss:102.62,     validation_loss:16.36\n",
      "training_loss:102.66,     validation_loss:15.99\n",
      "training_loss:102.63,     validation_loss:16.07\n",
      "training_loss:102.59,     validation_loss:16.60\n",
      "training_loss:102.48,     validation_loss:16.09\n",
      "training_loss:102.60,     validation_loss:15.89\n",
      "training_loss:102.71,     validation_loss:16.32\n",
      "alpha = 10, hidden layer = 6 seed = 48351\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.65,     validation_loss:16.38\n",
      "training_loss:102.58,     validation_loss:16.27\n",
      "training_loss:102.52,     validation_loss:16.45\n",
      "training_loss:102.65,     validation_loss:16.49\n",
      "alpha = 10, hidden layer = 6 seed = 19514\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.77,     validation_loss:15.99\n",
      "training_loss:102.59,     validation_loss:16.44\n",
      "training_loss:102.63,     validation_loss:16.12\n",
      "training_loss:102.64,     validation_loss:16.43\n",
      "alpha = 10, hidden layer = 6 seed = 68167\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.75,     validation_loss:16.02\n",
      "training_loss:102.62,     validation_loss:16.44\n",
      "training_loss:102.62,     validation_loss:16.11\n",
      "training_loss:102.56,     validation_loss:16.14\n",
      "training_loss:102.75,     validation_loss:15.81\n",
      "training_loss:102.77,     validation_loss:16.26\n",
      "alpha = 10, hidden layer = 6 seed = 48275\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.70,     validation_loss:16.30\n",
      "training_loss:102.63,     validation_loss:16.24\n",
      "training_loss:102.69,     validation_loss:16.42\n",
      "alpha = 10, hidden layer = 6 seed = 30650\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.68,     validation_loss:16.01\n",
      "training_loss:102.56,     validation_loss:16.55\n",
      "training_loss:102.57,     validation_loss:15.84\n",
      "training_loss:102.58,     validation_loss:16.60\n",
      "alpha = 10, hidden layer = 6 seed = 54546\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.62,     validation_loss:16.77\n",
      "training_loss:102.82,     validation_loss:16.10\n",
      "training_loss:102.60,     validation_loss:16.43\n",
      "training_loss:102.64,     validation_loss:15.85\n",
      "training_loss:102.72,     validation_loss:16.34\n",
      "alpha = 10, hidden layer = 6 seed = 4927\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.66,     validation_loss:16.49\n",
      "training_loss:102.71,     validation_loss:15.97\n",
      "training_loss:102.66,     validation_loss:16.03\n",
      "training_loss:102.60,     validation_loss:15.77\n",
      "training_loss:102.74,     validation_loss:15.81\n",
      "alpha = 10, hidden layer = 6 seed = 51683\n",
      "Hidden layer size of Refnet is  3771\n",
      "[2, 3771, 3771, 3771, 3771, 3771, 3771, 6129]\n",
      "training_loss:102.63,     validation_loss:16.10\n",
      "training_loss:102.75,     validation_loss:16.49\n",
      "alpha = 25, hidden layer = 1 seed = 76684\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.61,     validation_loss:16.58\n",
      "training_loss:102.51,     validation_loss:16.06\n",
      "training_loss:102.84,     validation_loss:16.06\n",
      "training_loss:102.43,     validation_loss:16.39\n",
      "training_loss:102.71,     validation_loss:15.74\n",
      "training_loss:102.59,     validation_loss:16.35\n",
      "training_loss:102.54,     validation_loss:16.26\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training_loss:102.67,     validation_loss:16.15\n",
      "training_loss:102.55,     validation_loss:16.24\n",
      "training_loss:102.49,     validation_loss:16.69\n",
      "training_loss:102.64,     validation_loss:15.99\n",
      "training_loss:102.65,     validation_loss:15.80\n",
      "training_loss:102.66,     validation_loss:16.02\n",
      "alpha = 25, hidden layer = 1 seed = 86728\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.63,     validation_loss:16.22\n",
      "training_loss:102.70,     validation_loss:16.14\n",
      "training_loss:102.64,     validation_loss:16.09\n",
      "training_loss:102.83,     validation_loss:16.23\n",
      "alpha = 25, hidden layer = 1 seed = 48351\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.79,     validation_loss:15.86\n",
      "training_loss:102.60,     validation_loss:16.31\n",
      "training_loss:102.66,     validation_loss:16.05\n",
      "training_loss:102.64,     validation_loss:16.36\n",
      "training_loss:102.52,     validation_loss:16.11\n",
      "training_loss:102.64,     validation_loss:16.13\n",
      "alpha = 25, hidden layer = 1 seed = 19514\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.64,     validation_loss:16.38\n",
      "training_loss:102.50,     validation_loss:16.18\n",
      "training_loss:102.74,     validation_loss:15.98\n",
      "training_loss:102.65,     validation_loss:16.14\n",
      "training_loss:102.68,     validation_loss:16.08\n",
      "training_loss:102.61,     validation_loss:15.97\n",
      "training_loss:102.72,     validation_loss:16.05\n",
      "alpha = 25, hidden layer = 1 seed = 68167\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.60,     validation_loss:16.09\n",
      "training_loss:102.75,     validation_loss:15.74\n",
      "training_loss:102.67,     validation_loss:16.19\n",
      "training_loss:102.61,     validation_loss:16.32\n",
      "training_loss:102.58,     validation_loss:16.05\n",
      "training_loss:102.54,     validation_loss:16.66\n",
      "training_loss:102.62,     validation_loss:15.97\n",
      "training_loss:102.57,     validation_loss:16.48\n",
      "training_loss:102.58,     validation_loss:16.21\n",
      "training_loss:102.56,     validation_loss:16.19\n",
      "training_loss:102.68,     validation_loss:16.20\n",
      "alpha = 25, hidden layer = 1 seed = 48275\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.62,     validation_loss:16.44\n",
      "training_loss:102.64,     validation_loss:16.37\n",
      "training_loss:102.58,     validation_loss:16.40\n",
      "training_loss:102.58,     validation_loss:16.32\n",
      "training_loss:102.55,     validation_loss:16.41\n",
      "training_loss:102.81,     validation_loss:16.15\n",
      "training_loss:102.47,     validation_loss:16.49\n",
      "training_loss:102.71,     validation_loss:16.30\n",
      "training_loss:102.59,     validation_loss:16.59\n",
      "training_loss:102.65,     validation_loss:15.85\n",
      "training_loss:102.62,     validation_loss:16.11\n",
      "training_loss:102.67,     validation_loss:15.82\n",
      "training_loss:102.60,     validation_loss:16.34\n",
      "training_loss:102.48,     validation_loss:16.04\n",
      "training_loss:102.70,     validation_loss:16.32\n",
      "alpha = 25, hidden layer = 1 seed = 30650\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.61,     validation_loss:16.26\n",
      "training_loss:102.62,     validation_loss:16.23\n",
      "training_loss:102.55,     validation_loss:16.02\n",
      "training_loss:102.60,     validation_loss:15.99\n",
      "training_loss:102.81,     validation_loss:16.24\n",
      "alpha = 25, hidden layer = 1 seed = 54546\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.66,     validation_loss:16.71\n",
      "training_loss:102.41,     validation_loss:16.17\n",
      "training_loss:102.60,     validation_loss:16.48\n",
      "alpha = 25, hidden layer = 1 seed = 4927\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.71,     validation_loss:15.70\n",
      "training_loss:102.70,     validation_loss:16.40\n",
      "training_loss:102.57,     validation_loss:16.39\n",
      "training_loss:102.60,     validation_loss:16.19\n",
      "training_loss:102.72,     validation_loss:16.19\n",
      "training_loss:102.53,     validation_loss:15.96\n",
      "training_loss:102.71,     validation_loss:16.08\n",
      "alpha = 25, hidden layer = 1 seed = 51683\n",
      "Hidden layer size of Refnet is  2971\n",
      "[2, 2971, 6129]\n",
      "training_loss:102.70,     validation_loss:16.51\n",
      "training_loss:102.59,     validation_loss:15.52\n",
      "training_loss:102.64,     validation_loss:16.46\n",
      "alpha = 25, hidden layer = 2 seed = 76684\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.74,     validation_loss:16.05\n",
      "training_loss:102.52,     validation_loss:16.40\n",
      "training_loss:102.67,     validation_loss:16.23\n",
      "training_loss:102.64,     validation_loss:16.19\n",
      "training_loss:102.55,     validation_loss:16.04\n",
      "training_loss:102.62,     validation_loss:16.27\n",
      "alpha = 25, hidden layer = 2 seed = 86728\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.62,     validation_loss:15.85\n",
      "training_loss:102.64,     validation_loss:16.56\n",
      "alpha = 25, hidden layer = 2 seed = 48351\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.72,     validation_loss:15.80\n",
      "training_loss:102.69,     validation_loss:16.31\n",
      "training_loss:102.66,     validation_loss:16.04\n",
      "training_loss:102.56,     validation_loss:16.02\n",
      "training_loss:102.68,     validation_loss:16.24\n",
      "alpha = 25, hidden layer = 2 seed = 19514\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.61,     validation_loss:16.19\n",
      "training_loss:102.74,     validation_loss:16.12\n",
      "training_loss:102.58,     validation_loss:16.28\n",
      "training_loss:102.65,     validation_loss:16.24\n",
      "training_loss:102.56,     validation_loss:16.20\n",
      "training_loss:102.66,     validation_loss:16.12\n",
      "training_loss:102.60,     validation_loss:15.99\n",
      "training_loss:102.70,     validation_loss:16.35\n",
      "alpha = 25, hidden layer = 2 seed = 68167\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.61,     validation_loss:16.18\n",
      "training_loss:102.81,     validation_loss:16.03\n",
      "training_loss:102.42,     validation_loss:16.45\n",
      "training_loss:102.67,     validation_loss:16.28\n",
      "training_loss:102.65,     validation_loss:15.99\n",
      "training_loss:102.63,     validation_loss:16.21\n",
      "training_loss:102.63,     validation_loss:16.21\n",
      "training_loss:102.68,     validation_loss:16.29\n",
      "alpha = 25, hidden layer = 2 seed = 48275\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.73,     validation_loss:16.40\n",
      "training_loss:102.56,     validation_loss:16.06\n",
      "training_loss:102.66,     validation_loss:16.07\n",
      "alpha = 25, hidden layer = 2 seed = 30650\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.58,     validation_loss:15.98\n",
      "training_loss:102.81,     validation_loss:16.07\n",
      "alpha = 25, hidden layer = 2 seed = 54546\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.75,     validation_loss:16.21\n",
      "training_loss:102.67,     validation_loss:15.95\n",
      "training_loss:102.69,     validation_loss:16.09\n",
      "alpha = 25, hidden layer = 2 seed = 4927\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.66,     validation_loss:16.34\n",
      "training_loss:102.63,     validation_loss:16.50\n",
      "training_loss:102.69,     validation_loss:15.93\n",
      "training_loss:102.68,     validation_loss:16.49\n",
      "training_loss:102.59,     validation_loss:16.15\n",
      "training_loss:102.60,     validation_loss:16.18\n",
      "alpha = 25, hidden layer = 2 seed = 51683\n",
      "Hidden layer size of Refnet is  4020\n",
      "[2, 4020, 4020, 6129]\n",
      "training_loss:102.65,     validation_loss:16.21\n",
      "training_loss:102.82,     validation_loss:16.21\n",
      "alpha = 25, hidden layer = 3 seed = 76684\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.66,     validation_loss:16.14\n",
      "training_loss:102.65,     validation_loss:16.33\n",
      "training_loss:102.66,     validation_loss:15.80\n",
      "training_loss:102.56,     validation_loss:16.13\n",
      "training_loss:102.61,     validation_loss:16.26\n",
      "alpha = 25, hidden layer = 3 seed = 86728\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.63,     validation_loss:16.25\n",
      "training_loss:102.63,     validation_loss:16.10\n",
      "training_loss:102.61,     validation_loss:16.07\n",
      "training_loss:102.63,     validation_loss:16.16\n",
      "alpha = 25, hidden layer = 3 seed = 48351\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.60,     validation_loss:15.98\n",
      "training_loss:102.72,     validation_loss:16.18\n",
      "alpha = 25, hidden layer = 3 seed = 19514\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.58,     validation_loss:16.33\n",
      "training_loss:102.68,     validation_loss:16.14\n",
      "training_loss:102.57,     validation_loss:16.03\n",
      "training_loss:102.65,     validation_loss:16.02\n",
      "training_loss:102.66,     validation_loss:16.58\n",
      "alpha = 25, hidden layer = 3 seed = 68167\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.62,     validation_loss:16.16\n",
      "training_loss:102.82,     validation_loss:16.14\n",
      "training_loss:102.47,     validation_loss:16.23\n",
      "training_loss:102.70,     validation_loss:15.92\n",
      "training_loss:102.48,     validation_loss:15.71\n",
      "training_loss:102.74,     validation_loss:16.37\n",
      "alpha = 25, hidden layer = 3 seed = 48275\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.68,     validation_loss:15.90\n",
      "training_loss:102.66,     validation_loss:16.03\n",
      "training_loss:102.57,     validation_loss:16.66\n",
      "training_loss:102.53,     validation_loss:16.62\n",
      "training_loss:102.62,     validation_loss:15.84\n",
      "training_loss:102.61,     validation_loss:16.62\n",
      "training_loss:102.55,     validation_loss:16.04\n",
      "training_loss:102.56,     validation_loss:16.57\n",
      "alpha = 25, hidden layer = 3 seed = 30650\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.68,     validation_loss:16.43\n",
      "training_loss:102.61,     validation_loss:16.38\n",
      "training_loss:102.63,     validation_loss:16.46\n",
      "alpha = 25, hidden layer = 3 seed = 54546\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.66,     validation_loss:15.79\n",
      "training_loss:102.65,     validation_loss:16.58\n",
      "training_loss:102.60,     validation_loss:16.18\n",
      "training_loss:102.50,     validation_loss:16.27\n",
      "training_loss:102.69,     validation_loss:16.12\n",
      "training_loss:102.46,     validation_loss:16.37\n",
      "training_loss:102.57,     validation_loss:16.08\n",
      "training_loss:102.61,     validation_loss:16.29\n",
      "alpha = 25, hidden layer = 3 seed = 4927\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.66,     validation_loss:16.15\n",
      "training_loss:102.63,     validation_loss:16.28\n",
      "training_loss:102.64,     validation_loss:16.18\n",
      "training_loss:102.57,     validation_loss:16.00\n",
      "training_loss:102.65,     validation_loss:16.17\n",
      "alpha = 25, hidden layer = 3 seed = 51683\n",
      "Hidden layer size of Refnet is  4970\n",
      "[2, 4970, 4970, 4970, 6129]\n",
      "training_loss:102.59,     validation_loss:16.47\n",
      "training_loss:102.61,     validation_loss:16.19\n",
      "training_loss:102.65,     validation_loss:15.92\n",
      "training_loss:102.51,     validation_loss:16.23\n",
      "training_loss:102.67,     validation_loss:16.27\n",
      "alpha = 25, hidden layer = 4 seed = 76684\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.54,     validation_loss:16.53\n",
      "training_loss:102.63,     validation_loss:16.12\n",
      "training_loss:102.57,     validation_loss:16.03\n",
      "training_loss:102.73,     validation_loss:16.23\n",
      "alpha = 25, hidden layer = 4 seed = 86728\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.65,     validation_loss:16.12\n",
      "training_loss:102.64,     validation_loss:16.00\n",
      "training_loss:102.64,     validation_loss:16.29\n",
      "alpha = 25, hidden layer = 4 seed = 48351\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.65,     validation_loss:16.16\n",
      "training_loss:102.52,     validation_loss:16.02\n",
      "training_loss:102.75,     validation_loss:16.14\n",
      "alpha = 25, hidden layer = 4 seed = 19514\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.84,     validation_loss:16.45\n",
      "training_loss:102.52,     validation_loss:16.05\n",
      "training_loss:102.71,     validation_loss:16.59\n",
      "alpha = 25, hidden layer = 4 seed = 68167\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.62,     validation_loss:16.40\n",
      "training_loss:102.58,     validation_loss:16.68\n",
      "training_loss:102.48,     validation_loss:16.42\n",
      "training_loss:102.71,     validation_loss:16.64\n",
      "alpha = 25, hidden layer = 4 seed = 48275\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.64,     validation_loss:16.10\n",
      "training_loss:102.59,     validation_loss:16.23\n",
      "training_loss:102.67,     validation_loss:15.82\n",
      "training_loss:102.77,     validation_loss:16.32\n",
      "alpha = 25, hidden layer = 4 seed = 30650\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.68,     validation_loss:16.34\n",
      "training_loss:102.55,     validation_loss:16.74\n",
      "training_loss:102.58,     validation_loss:16.22\n",
      "training_loss:102.61,     validation_loss:16.45\n",
      "alpha = 25, hidden layer = 4 seed = 54546\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.68,     validation_loss:16.13\n",
      "training_loss:102.65,     validation_loss:16.14\n",
      "training_loss:102.69,     validation_loss:15.93\n",
      "training_loss:102.68,     validation_loss:16.36\n",
      "training_loss:102.66,     validation_loss:16.63\n",
      "training_loss:102.58,     validation_loss:16.35\n",
      "training_loss:102.56,     validation_loss:16.58\n",
      "training_loss:102.73,     validation_loss:16.48\n",
      "training_loss:102.61,     validation_loss:15.52\n",
      "training_loss:102.69,     validation_loss:16.07\n",
      "alpha = 25, hidden layer = 4 seed = 4927\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.64,     validation_loss:16.41\n",
      "training_loss:102.71,     validation_loss:16.31\n",
      "training_loss:102.64,     validation_loss:16.40\n",
      "training_loss:102.56,     validation_loss:16.17\n",
      "training_loss:102.63,     validation_loss:16.26\n",
      "alpha = 25, hidden layer = 4 seed = 51683\n",
      "Hidden layer size of Refnet is  5844\n",
      "[2, 5844, 5844, 5844, 5844, 6129]\n",
      "training_loss:102.71,     validation_loss:16.16\n",
      "training_loss:102.59,     validation_loss:16.29\n",
      "training_loss:102.69,     validation_loss:15.78\n",
      "training_loss:102.72,     validation_loss:16.36\n",
      "alpha = 25, hidden layer = 5 seed = 76684\n",
      "Hidden layer size of Refnet is  6659\n",
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.71,     validation_loss:15.94\n",
      "training_loss:102.71,     validation_loss:16.42\n",
      "training_loss:102.59,     validation_loss:16.21\n",
      "training_loss:102.59,     validation_loss:16.36\n",
      "training_loss:102.71,     validation_loss:15.97\n",
      "training_loss:102.64,     validation_loss:16.15\n",
      "training_loss:102.66,     validation_loss:16.14\n",
      "training_loss:102.50,     validation_loss:16.47\n",
      "training_loss:102.66,     validation_loss:16.29\n",
      "training_loss:102.63,     validation_loss:16.08\n",
      "training_loss:102.48,     validation_loss:16.44\n",
      "training_loss:102.77,     validation_loss:15.61\n",
      "training_loss:102.59,     validation_loss:16.30\n",
      "training_loss:102.59,     validation_loss:16.29\n",
      "training_loss:102.76,     validation_loss:16.11\n",
      "training_loss:102.51,     validation_loss:16.29\n",
      "training_loss:102.57,     validation_loss:16.62\n",
      "alpha = 25, hidden layer = 5 seed = 86728\n",
      "Hidden layer size of Refnet is  6659\n",
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.67,     validation_loss:16.13\n",
      "training_loss:102.58,     validation_loss:16.13\n",
      "training_loss:102.61,     validation_loss:15.87\n",
      "training_loss:102.58,     validation_loss:16.46\n",
      "training_loss:102.61,     validation_loss:16.04\n",
      "training_loss:102.65,     validation_loss:15.74\n",
      "training_loss:102.59,     validation_loss:16.07\n",
      "training_loss:102.64,     validation_loss:15.99\n",
      "training_loss:102.56,     validation_loss:16.10\n",
      "training_loss:102.67,     validation_loss:16.64\n",
      "alpha = 25, hidden layer = 5 seed = 48351\n",
      "Hidden layer size of Refnet is  6659\n",
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.63,     validation_loss:16.12\n",
      "training_loss:102.65,     validation_loss:16.14\n",
      "alpha = 25, hidden layer = 5 seed = 19514\n",
      "Hidden layer size of Refnet is  6659\n",
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.67,     validation_loss:16.09\n",
      "training_loss:102.64,     validation_loss:16.26\n",
      "training_loss:102.76,     validation_loss:16.46\n",
      "alpha = 25, hidden layer = 5 seed = 68167\n",
      "Hidden layer size of Refnet is  6659\n",
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.72,     validation_loss:16.23\n",
      "training_loss:102.59,     validation_loss:15.86\n",
      "training_loss:102.66,     validation_loss:16.25\n",
      "alpha = 25, hidden layer = 5 seed = 48275\n",
      "Hidden layer size of Refnet is  6659\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.68,     validation_loss:15.93\n",
      "training_loss:102.52,     validation_loss:16.54\n",
      "training_loss:102.67,     validation_loss:16.08\n",
      "training_loss:102.57,     validation_loss:16.26\n",
      "training_loss:102.70,     validation_loss:16.21\n",
      "training_loss:102.72,     validation_loss:16.11\n",
      "training_loss:102.53,     validation_loss:16.38\n",
      "training_loss:102.55,     validation_loss:15.88\n",
      "training_loss:102.66,     validation_loss:16.13\n",
      "alpha = 25, hidden layer = 5 seed = 30650\n",
      "Hidden layer size of Refnet is  6659\n",
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.68,     validation_loss:15.95\n",
      "training_loss:102.65,     validation_loss:16.22\n",
      "training_loss:102.65,     validation_loss:16.16\n",
      "training_loss:102.39,     validation_loss:16.25\n",
      "training_loss:102.84,     validation_loss:15.83\n",
      "training_loss:102.62,     validation_loss:16.48\n",
      "training_loss:102.55,     validation_loss:16.10\n",
      "training_loss:102.71,     validation_loss:16.11\n",
      "alpha = 25, hidden layer = 5 seed = 54546\n",
      "Hidden layer size of Refnet is  6659\n",
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.66,     validation_loss:15.98\n",
      "training_loss:102.69,     validation_loss:16.04\n",
      "alpha = 25, hidden layer = 5 seed = 4927\n",
      "Hidden layer size of Refnet is  6659\n",
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.63,     validation_loss:16.46\n",
      "training_loss:102.56,     validation_loss:16.52\n",
      "training_loss:102.61,     validation_loss:16.23\n",
      "training_loss:102.64,     validation_loss:16.24\n",
      "alpha = 25, hidden layer = 5 seed = 51683\n",
      "Hidden layer size of Refnet is  6659\n",
      "[2, 6659, 6659, 6659, 6659, 6659, 6129]\n",
      "training_loss:102.64,     validation_loss:16.63\n",
      "training_loss:102.62,     validation_loss:16.34\n",
      "training_loss:102.51,     validation_loss:15.97\n",
      "training_loss:102.63,     validation_loss:16.25\n",
      "alpha = 25, hidden layer = 6 seed = 76684\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.71,     validation_loss:16.07\n",
      "training_loss:102.58,     validation_loss:16.60\n",
      "training_loss:102.61,     validation_loss:16.00\n",
      "training_loss:102.52,     validation_loss:16.31\n",
      "training_loss:102.62,     validation_loss:16.02\n",
      "training_loss:102.58,     validation_loss:15.97\n",
      "training_loss:102.71,     validation_loss:16.23\n",
      "alpha = 25, hidden layer = 6 seed = 86728\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.66,     validation_loss:16.26\n",
      "training_loss:102.69,     validation_loss:16.58\n",
      "alpha = 25, hidden layer = 6 seed = 48351\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.72,     validation_loss:15.91\n",
      "training_loss:102.78,     validation_loss:15.87\n",
      "training_loss:102.51,     validation_loss:16.48\n",
      "training_loss:102.56,     validation_loss:16.24\n",
      "training_loss:102.61,     validation_loss:16.13\n",
      "training_loss:102.71,     validation_loss:16.09\n",
      "training_loss:102.54,     validation_loss:16.25\n",
      "training_loss:102.72,     validation_loss:16.21\n",
      "training_loss:102.67,     validation_loss:16.25\n",
      "training_loss:102.55,     validation_loss:16.19\n",
      "training_loss:102.75,     validation_loss:16.00\n",
      "training_loss:102.68,     validation_loss:16.50\n",
      "training_loss:102.44,     validation_loss:16.44\n",
      "training_loss:102.64,     validation_loss:16.28\n",
      "training_loss:102.71,     validation_loss:16.06\n",
      "training_loss:102.58,     validation_loss:15.74\n",
      "training_loss:102.67,     validation_loss:16.19\n",
      "alpha = 25, hidden layer = 6 seed = 19514\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.61,     validation_loss:16.57\n",
      "training_loss:102.58,     validation_loss:16.25\n",
      "training_loss:102.62,     validation_loss:15.95\n",
      "training_loss:102.64,     validation_loss:16.58\n",
      "alpha = 25, hidden layer = 6 seed = 68167\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.69,     validation_loss:16.39\n",
      "training_loss:102.65,     validation_loss:16.09\n",
      "training_loss:102.63,     validation_loss:16.06\n",
      "training_loss:102.68,     validation_loss:16.11\n",
      "alpha = 25, hidden layer = 6 seed = 48275\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.57,     validation_loss:16.86\n",
      "training_loss:102.56,     validation_loss:16.29\n",
      "training_loss:102.58,     validation_loss:16.00\n",
      "training_loss:102.75,     validation_loss:16.22\n",
      "alpha = 25, hidden layer = 6 seed = 30650\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.70,     validation_loss:15.80\n",
      "training_loss:102.70,     validation_loss:16.00\n",
      "training_loss:102.60,     validation_loss:16.35\n",
      "training_loss:102.45,     validation_loss:16.47\n",
      "training_loss:102.65,     validation_loss:16.43\n",
      "training_loss:102.58,     validation_loss:16.02\n",
      "training_loss:102.53,     validation_loss:16.23\n",
      "training_loss:102.64,     validation_loss:16.55\n",
      "alpha = 25, hidden layer = 6 seed = 54546\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.75,     validation_loss:16.42\n",
      "training_loss:102.57,     validation_loss:15.86\n",
      "training_loss:102.70,     validation_loss:16.48\n",
      "alpha = 25, hidden layer = 6 seed = 4927\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.63,     validation_loss:16.58\n",
      "training_loss:102.60,     validation_loss:16.19\n",
      "training_loss:102.59,     validation_loss:15.98\n",
      "training_loss:102.65,     validation_loss:16.09\n",
      "alpha = 25, hidden layer = 6 seed = 51683\n",
      "Hidden layer size of Refnet is  7425\n",
      "[2, 7425, 7425, 7425, 7425, 7425, 7425, 6129]\n",
      "training_loss:102.67,     validation_loss:16.26\n",
      "training_loss:102.51,     validation_loss:16.31\n",
      "training_loss:102.75,     validation_loss:15.88\n",
      "training_loss:102.68,     validation_loss:16.22\n",
      "training_loss:102.58,     validation_loss:16.18\n",
      "training_loss:102.64,     validation_loss:16.07\n",
      "training_loss:102.64,     validation_loss:16.25\n",
      "alpha = 50, hidden layer = 1 seed = 76684\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.60,     validation_loss:16.25\n",
      "training_loss:102.60,     validation_loss:16.30\n",
      "training_loss:102.68,     validation_loss:16.23\n",
      "training_loss:102.68,     validation_loss:16.16\n",
      "training_loss:102.53,     validation_loss:15.88\n",
      "training_loss:102.59,     validation_loss:16.21\n",
      "alpha = 50, hidden layer = 1 seed = 86728\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.67,     validation_loss:16.60\n",
      "training_loss:102.65,     validation_loss:16.55\n",
      "training_loss:102.53,     validation_loss:15.99\n",
      "training_loss:102.75,     validation_loss:15.99\n",
      "alpha = 50, hidden layer = 1 seed = 48351\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.61,     validation_loss:16.45\n",
      "training_loss:102.59,     validation_loss:16.05\n",
      "training_loss:102.69,     validation_loss:16.50\n",
      "alpha = 50, hidden layer = 1 seed = 19514\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.69,     validation_loss:15.94\n",
      "training_loss:102.69,     validation_loss:16.15\n",
      "training_loss:102.69,     validation_loss:16.24\n",
      "alpha = 50, hidden layer = 1 seed = 68167\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.67,     validation_loss:16.28\n",
      "training_loss:102.51,     validation_loss:16.51\n",
      "training_loss:102.58,     validation_loss:16.02\n",
      "training_loss:102.72,     validation_loss:16.08\n",
      "alpha = 50, hidden layer = 1 seed = 48275\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.64,     validation_loss:16.14\n",
      "training_loss:102.71,     validation_loss:16.37\n",
      "alpha = 50, hidden layer = 1 seed = 30650\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.70,     validation_loss:16.40\n",
      "training_loss:102.59,     validation_loss:16.29\n",
      "training_loss:102.62,     validation_loss:16.02\n",
      "training_loss:102.68,     validation_loss:16.06\n",
      "alpha = 50, hidden layer = 1 seed = 54546\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.63,     validation_loss:16.78\n",
      "training_loss:102.61,     validation_loss:15.75\n",
      "training_loss:102.67,     validation_loss:16.45\n",
      "alpha = 50, hidden layer = 1 seed = 4927\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.66,     validation_loss:15.79\n",
      "training_loss:102.80,     validation_loss:16.12\n",
      "alpha = 50, hidden layer = 1 seed = 51683\n",
      "Hidden layer size of Refnet is  3234\n",
      "[2, 3234, 6129]\n",
      "training_loss:102.69,     validation_loss:16.08\n",
      "training_loss:102.65,     validation_loss:16.31\n",
      "training_loss:102.61,     validation_loss:15.94\n",
      "training_loss:102.75,     validation_loss:16.17\n",
      "alpha = 50, hidden layer = 2 seed = 76684\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.65,     validation_loss:16.32\n",
      "training_loss:102.70,     validation_loss:16.30\n",
      "training_loss:102.51,     validation_loss:16.29\n",
      "training_loss:102.75,     validation_loss:16.14\n",
      "training_loss:102.53,     validation_loss:15.95\n",
      "training_loss:102.64,     validation_loss:16.44\n",
      "alpha = 50, hidden layer = 2 seed = 86728\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.64,     validation_loss:16.77\n",
      "training_loss:102.56,     validation_loss:16.36\n",
      "training_loss:102.61,     validation_loss:15.55\n",
      "training_loss:102.78,     validation_loss:16.30\n",
      "alpha = 50, hidden layer = 2 seed = 48351\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.74,     validation_loss:16.30\n",
      "training_loss:102.64,     validation_loss:16.08\n",
      "training_loss:102.63,     validation_loss:15.83\n",
      "training_loss:102.70,     validation_loss:16.10\n",
      "alpha = 50, hidden layer = 2 seed = 19514\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.72,     validation_loss:16.14\n",
      "training_loss:102.66,     validation_loss:16.51\n",
      "training_loss:102.59,     validation_loss:16.14\n",
      "training_loss:102.70,     validation_loss:16.34\n",
      "alpha = 50, hidden layer = 2 seed = 68167\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.64,     validation_loss:16.20\n",
      "training_loss:102.57,     validation_loss:15.87\n",
      "training_loss:102.71,     validation_loss:16.02\n",
      "alpha = 50, hidden layer = 2 seed = 48275\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.57,     validation_loss:16.45\n",
      "training_loss:102.64,     validation_loss:16.36\n",
      "training_loss:102.60,     validation_loss:16.39\n",
      "training_loss:102.53,     validation_loss:16.35\n",
      "training_loss:102.65,     validation_loss:16.19\n",
      "training_loss:102.53,     validation_loss:16.41\n",
      "training_loss:102.63,     validation_loss:15.85\n",
      "training_loss:102.66,     validation_loss:16.37\n",
      "alpha = 50, hidden layer = 2 seed = 30650\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.64,     validation_loss:16.40\n",
      "training_loss:102.68,     validation_loss:16.59\n",
      "alpha = 50, hidden layer = 2 seed = 54546\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.68,     validation_loss:16.07\n",
      "training_loss:102.73,     validation_loss:16.24\n",
      "alpha = 50, hidden layer = 2 seed = 4927\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.71,     validation_loss:16.24\n",
      "training_loss:102.58,     validation_loss:15.96\n",
      "training_loss:102.60,     validation_loss:16.13\n",
      "alpha = 50, hidden layer = 2 seed = 51683\n",
      "Hidden layer size of Refnet is  4486\n",
      "[2, 4486, 4486, 6129]\n",
      "training_loss:102.67,     validation_loss:16.22\n",
      "training_loss:102.62,     validation_loss:16.36\n",
      "training_loss:102.52,     validation_loss:16.27\n",
      "training_loss:102.60,     validation_loss:16.21\n",
      "training_loss:102.66,     validation_loss:16.51\n",
      "alpha = 50, hidden layer = 3 seed = 76684\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.71,     validation_loss:16.86\n",
      "training_loss:102.48,     validation_loss:16.29\n",
      "training_loss:102.72,     validation_loss:16.12\n",
      "training_loss:102.55,     validation_loss:16.43\n",
      "training_loss:102.66,     validation_loss:15.87\n",
      "training_loss:102.55,     validation_loss:16.34\n",
      "training_loss:102.64,     validation_loss:16.36\n",
      "alpha = 50, hidden layer = 3 seed = 86728\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.65,     validation_loss:16.47\n",
      "training_loss:102.71,     validation_loss:16.32\n",
      "training_loss:102.54,     validation_loss:16.20\n",
      "training_loss:102.63,     validation_loss:16.36\n",
      "alpha = 50, hidden layer = 3 seed = 48351\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.63,     validation_loss:16.17\n",
      "training_loss:102.63,     validation_loss:15.99\n",
      "training_loss:102.65,     validation_loss:16.40\n",
      "alpha = 50, hidden layer = 3 seed = 19514\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.68,     validation_loss:16.57\n",
      "training_loss:102.66,     validation_loss:16.09\n",
      "training_loss:102.57,     validation_loss:16.16\n",
      "training_loss:102.58,     validation_loss:16.18\n",
      "alpha = 50, hidden layer = 3 seed = 68167\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.63,     validation_loss:16.57\n",
      "training_loss:102.57,     validation_loss:16.37\n",
      "training_loss:102.50,     validation_loss:16.30\n",
      "training_loss:102.69,     validation_loss:16.45\n",
      "alpha = 50, hidden layer = 3 seed = 48275\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.64,     validation_loss:16.39\n",
      "training_loss:102.59,     validation_loss:16.66\n",
      "training_loss:102.46,     validation_loss:16.27\n",
      "training_loss:102.64,     validation_loss:16.22\n",
      "training_loss:102.58,     validation_loss:16.10\n",
      "training_loss:102.67,     validation_loss:15.95\n",
      "training_loss:102.57,     validation_loss:16.09\n",
      "training_loss:102.57,     validation_loss:16.34\n",
      "alpha = 50, hidden layer = 3 seed = 30650\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.57,     validation_loss:16.63\n",
      "training_loss:102.54,     validation_loss:16.38\n",
      "training_loss:102.68,     validation_loss:16.00\n",
      "training_loss:102.61,     validation_loss:15.89\n",
      "training_loss:102.55,     validation_loss:16.52\n",
      "training_loss:102.52,     validation_loss:16.10\n",
      "training_loss:102.67,     validation_loss:16.37\n",
      "alpha = 50, hidden layer = 3 seed = 54546\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.62,     validation_loss:15.97\n",
      "training_loss:102.75,     validation_loss:16.47\n",
      "alpha = 50, hidden layer = 3 seed = 4927\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.72,     validation_loss:15.87\n",
      "training_loss:102.64,     validation_loss:15.96\n",
      "training_loss:102.77,     validation_loss:16.04\n",
      "alpha = 50, hidden layer = 3 seed = 51683\n",
      "Hidden layer size of Refnet is  5606\n",
      "[2, 5606, 5606, 5606, 6129]\n",
      "training_loss:102.70,     validation_loss:16.28\n",
      "training_loss:102.73,     validation_loss:16.35\n",
      "alpha = 50, hidden layer = 4 seed = 76684\n",
      "Hidden layer size of Refnet is  6628\n",
      "[2, 6628, 6628, 6628, 6628, 6129]\n",
      "training_loss:102.64,     validation_loss:16.49\n",
      "training_loss:102.64,     validation_loss:16.06\n",
      "training_loss:102.56,     validation_loss:16.15\n",
      "training_loss:102.63,     validation_loss:16.09\n",
      "training_loss:102.72,     validation_loss:16.36\n",
      "alpha = 50, hidden layer = 4 seed = 86728\n",
      "Hidden layer size of Refnet is  6628\n",
      "[2, 6628, 6628, 6628, 6628, 6129]\n",
      "training_loss:102.73,     validation_loss:16.30\n",
      "training_loss:102.67,     validation_loss:16.09\n",
      "training_loss:102.73,     validation_loss:16.37\n",
      "alpha = 50, hidden layer = 4 seed = 48351\n",
      "Hidden layer size of Refnet is  6628\n",
      "[2, 6628, 6628, 6628, 6628, 6129]\n",
      "training_loss:102.69,     validation_loss:16.01\n",
      "training_loss:102.76,     validation_loss:16.15\n",
      "alpha = 50, hidden layer = 4 seed = 19514\n",
      "Hidden layer size of Refnet is  6628\n",
      "[2, 6628, 6628, 6628, 6628, 6129]\n",
      "training_loss:102.67,     validation_loss:16.37\n",
      "training_loss:102.57,     validation_loss:15.93\n",
      "training_loss:102.66,     validation_loss:16.28\n",
      "alpha = 50, hidden layer = 4 seed = 68167\n",
      "Hidden layer size of Refnet is  6628\n",
      "[2, 6628, 6628, 6628, 6628, 6129]\n",
      "training_loss:102.64,     validation_loss:16.24\n",
      "training_loss:102.59,     validation_loss:16.21\n",
      "training_loss:102.73,     validation_loss:16.38\n",
      "alpha = 50, hidden layer = 4 seed = 48275\n",
      "Hidden layer size of Refnet is  6628\n",
      "[2, 6628, 6628, 6628, 6628, 6129]\n",
      "training_loss:102.78,     validation_loss:16.06\n",
      "training_loss:102.63,     validation_loss:16.20\n",
      "training_loss:102.67,     validation_loss:16.10\n",
      "training_loss:102.55,     validation_loss:16.24\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training_loss:102.79,     validation_loss:16.13\n",
      "training_loss:102.60,     validation_loss:16.10\n",
      "training_loss:102.64,     validation_loss:16.30\n",
      "alpha = 50, hidden layer = 4 seed = 30650\n",
      "Hidden layer size of Refnet is  6628\n",
      "[2, 6628, 6628, 6628, 6628, 6129]\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-185-009237f11152>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m     19\u001b[0m             \u001b[1;31m#r2_score = r2score(model)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     20\u001b[0m             \u001b[0moptimizer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0moptim\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mAdam\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mparameters\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlr\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m6.89E-04\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mweight_decay\u001b[0m\u001b[1;33m=\u001b[0m \u001b[1;36m1e-1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 21\u001b[1;33m             \u001b[0mtotal_loss\u001b[0m\u001b[1;33m=\u001b[0m \u001b[0mtrain\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mtrain_loader\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mval_loader\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0moptimizer\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     22\u001b[0m             \u001b[1;32mfor\u001b[0m \u001b[0mtraining_loss\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mvalidation_loss\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mzip\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtotal_loss\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'training_loss'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mtotal_loss\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'validation_loss'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     23\u001b[0m                 \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m'training_loss:{:.2f},     validation_loss:{:.2f}'\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtraining_loss\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mvalidation_loss\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m<ipython-input-170-f063679c61c1>\u001b[0m in \u001b[0;36mtrain\u001b[1;34m(model, train_loader, val_loader, optimizer, epochs)\u001b[0m\n\u001b[0;32m     16\u001b[0m             \u001b[0myhat\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mx\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     17\u001b[0m             \u001b[0mloss\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mMSE\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0my\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0myhat\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 18\u001b[1;33m             \u001b[0mloss\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     19\u001b[0m             \u001b[0moptimizer\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mstep\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     20\u001b[0m             \u001b[0mbatch_loss\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mloss\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\ProgramData\\Anaconda3\\lib\\site-packages\\torch\\tensor.py\u001b[0m in \u001b[0;36mbackward\u001b[1;34m(self, gradient, retain_graph, create_graph)\u001b[0m\n\u001b[0;32m    183\u001b[0m                 \u001b[0mproducts\u001b[0m\u001b[1;33m.\u001b[0m \u001b[0mDefaults\u001b[0m \u001b[0mto\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;31m`\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[0;31m`\u001b[0m\u001b[0;31m`\u001b[0m\u001b[1;33m.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    184\u001b[0m         \"\"\"\n\u001b[1;32m--> 185\u001b[1;33m         \u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mautograd\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mgradient\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    186\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    187\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0mregister_hook\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mhook\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mC:\\ProgramData\\Anaconda3\\lib\\site-packages\\torch\\autograd\\__init__.py\u001b[0m in \u001b[0;36mbackward\u001b[1;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables)\u001b[0m\n\u001b[0;32m    125\u001b[0m     Variable._execution_engine.run_backward(\n\u001b[0;32m    126\u001b[0m         \u001b[0mtensors\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mgrad_tensors\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 127\u001b[1;33m         allow_unreachable=True)  # allow_unreachable flag\n\u001b[0m\u001b[0;32m    128\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    129\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "start = torch.cuda.Event(enable_timing=True)\n",
    "end = torch.cuda.Event(enable_timing=True)\n",
    "start.record()\n",
    "\n",
    "in_size = 2\n",
    "d = 2043*3     #depth of hidden layer\n",
    "seed_list = np.random.randint(60,99999,10)\n",
    "print(\"random seeds are = {}\".format(seed_list))\n",
    "alpha_list = [1,2.5,5,10,25,50]\n",
    "#h =  number of hidden layers\n",
    "\n",
    "for alpha in alpha_list:\n",
    "    for h in range (1,7,1):\n",
    "        for seed in seed_list:\n",
    "            torch.manual_seed(seed)Helmholtz-Zentrum Dresden Rossendorf, Bautzner Landstraße 400, 01328 Dresden\n",
    "            print(\"alpha = {}, hidden layer = {} seed = {}\" .format(alpha,h,seed))\n",
    "            MSE = nn.MSELoss()\n",
    "            model = My_Model(in_size, alpha, h,d, p=0.20)\n",
    "            #r2_score = r2score(model)\n",
    "            optimizer = optim.Adam(model.parameters(), lr=6.89E-04, weight_decay= 1e-1)\n",
    "            total_loss= train(model,train_loader, val_loader, optimizer)\n",
    "            for training_loss,validation_loss in zip(total_loss['training_loss'],total_loss['validation_loss']):\n",
    "                print('training_loss:{:.2f},     validation_loss:{:.2f}'.format(training_loss,validation_loss)) \n",
    "end.record()\n",
    "torch.cuda.synchronize()\n",
    "print(\"time taken to run this program is {:0.4f} seconds\" .format(start.elapsed_time(end)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Calculate R squared score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8579488678207335\n"
     ]
    }
   ],
   "source": [
    "\n",
    "from sklearn.metrics import r2_score\n",
    "\n",
    "def r2score(model):\n",
    "    x = data[:][0][0]#work on the indexes\n",
    "    y = data[:][1][0]\n",
    "    yhat= model(x)\n",
    "    yhat = yhat.detach().numpy()\n",
    "    \n",
    "    \n",
    "    r2_score22 = r2_score(y,yhat)\n",
    "    #print(MSE)\n",
    "    print(r2_score22)\n",
    "    #print(\"R2 score  = {}\".format(r2_score_cal))\n",
    "\n",
    "r2_score = r2score(model)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2 score  = 0.01471201702952385\n"
     ]
    }
   ],
   "source": [
    "def r2score(model):\n",
    "    x = data[:][0][0]\n",
    "    y = data[:][1][0]\n",
    "    yhat= model(x)\n",
    "    yhat = yhat.detach().numpy()\n",
    "    \n",
    "    for i in range (len(y)):\n",
    "        MSE = ((y[i]-yhat[i])**2).sum()\n",
    "    for i in range (len(data)):\n",
    "        tss = (((data[:][1][i] - data[:][1].mean())**2).sum())#index issue\n",
    "    r2_score_cal = (tss - MSE)/tss\n",
    "    \n",
    "    \n",
    "    #print(MSE)\n",
    "    print(\"R2 score  = {}\".format(r2_score_cal))\n",
    "\n",
    "r2_score = r2score(model)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "useless LRfinder find the steepest decent\n",
    "issue with r2 score\n",
    "add files to cudo server\n",
    "issue with the input and y values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_default_device():\n",
    "    \"\"\"Pick GPU if available, else CPU\"\"\"\n",
    "    if torch.cuda.is_available():\n",
    "        return torch.device('cuda')\n",
    "    else:\n",
    "        return torch.device('cpu')\n",
    "\n",
    "device = get_default_device()\n",
    "print(device)\n",
    "\n",
    "def to_device(data, device):\n",
    "    \"\"\"Move tensor(s) to chosen device\"\"\"\n",
    "    if isinstance(data, (list,tuple)):\n",
    "        return [to_device(x, device) for x,y in data]\n",
    "    return data.to(device, non_blocking=True)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}