cross_verifcation.py.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "source": [
    "###该代码用于通过k折交叉验证测试模型的精度###\n",
    "将数据集（lenths.xlsx）读入后进行模型训练并评估\n",
    "trainModel=False为不需要重新训练，直接从net/last.pt读取模型"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%% md\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch import nn\n",
    "import torch.nn.functional as F\n",
    "import readExcel\n",
    "import time"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "outputs": [],
   "source": [
    "# Define model\n",
    "class Net(nn.Module):\n",
    "    def __init__(self, net_frame):\n",
    "        super(Net, self).__init__()\n",
    "        n_input, n_hidden1, n_output = net_frame\n",
    "        self.net = nn.Sequential(\n",
    "                    nn.Linear(n_input, n_hidden1),\n",
    "                    nn.ReLU(),\n",
    "                    nn.Linear(n_hidden1, n_output),\n",
    "                    )\n",
    "\n",
    "    def forward(self, input):\n",
    "        return self.net(input)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "outputs": [],
   "source": [
    "# 设置超参数\n",
    "learning_rate = 0.0001  # 学习率\n",
    "net_frame = [7, 3, 6]  # 各层感知机数量\n",
    "iteration = 150000  # 迭代次数\n",
    "k_cross=5\n",
    "trainModel = True"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 51 entries, 0 to 50\n",
      "Data columns (total 14 columns):\n",
      " #   Column  Non-Null Count  Dtype  \n",
      "---  ------  --------------  -----  \n",
      " 0   ID      51 non-null     int64  \n",
      " 1   身高      51 non-null     float64\n",
      " 2   肩宽      51 non-null     float64\n",
      " 3   臀宽      51 non-null     float64\n",
      " 4   大臂      51 non-null     int64  \n",
      " 5   小臂      51 non-null     int64  \n",
      " 6   大腿      51 non-null     int64  \n",
      " 7   小腿      51 non-null     int64  \n",
      " 8   目测身高    51 non-null     int64  \n",
      " 9   目测肩宽    51 non-null     int64  \n",
      " 10  目测大臂长   51 non-null     int64  \n",
      " 11  目测小臂长   51 non-null     int64  \n",
      " 12  目测大腿长   51 non-null     int64  \n",
      " 13  目测小腿长   51 non-null     int64  \n",
      "dtypes: float64(3), int64(11)\n",
      "memory usage: 5.7 KB\n"
     ]
    },
    {
     "data": {
      "text/plain": "[[1.0,\n  411.0048661512416,\n  101.0,\n  112.2853507809456,\n  57.0,\n  48.0,\n  84.0,\n  78.0,\n  174.0,\n  47.0,\n  27.0,\n  24.0,\n  41.0,\n  37.0],\n [2.0,\n  411.0304125000971,\n  95.08417323613851,\n  107.1167587261676,\n  61.0,\n  51.0,\n  91.0,\n  83.0,\n  184.0,\n  46.0,\n  26.0,\n  24.0,\n  43.0,\n  38.0],\n [3.0,\n  368.0054347424777,\n  93.02150289046077,\n  130.0615239031129,\n  55.0,\n  45.0,\n  73.0,\n  68.0,\n  187.0,\n  47.0,\n  28.0,\n  26.0,\n  46.0,\n  38.0],\n [4.0,\n  417.1726261393478,\n  99.02019995940222,\n  104.0,\n  62.0,\n  56.0,\n  86.0,\n  84.0,\n  186.0,\n  48.0,\n  31.0,\n  27.0,\n  45.0,\n  37.0],\n [5.0,\n  386.0634662850138,\n  99.04544411531506,\n  119.016805536025,\n  60.0,\n  42.0,\n  79.0,\n  69.0,\n  188.0,\n  47.0,\n  30.0,\n  26.0,\n  42.0,\n  39.0],\n [6.0,\n  408.1482573771448,\n  95.13148795220224,\n  117.1067888723792,\n  62.0,\n  54.0,\n  91.0,\n  83.0,\n  174.0,\n  45.0,\n  27.0,\n  23.0,\n  40.0,\n  32.0],\n [7.0,\n  440.0727212632021,\n  107.0420478129973,\n  114.0,\n  62.0,\n  52.0,\n  95.0,\n  81.0,\n  162.0,\n  40.0,\n  30.0,\n  23.0,\n  37.0,\n  32.0],\n [8.0,\n  411.0048661512416,\n  99.02019995940222,\n  113.0,\n  62.0,\n  52.0,\n  77.0,\n  72.0,\n  164.0,\n  42.0,\n  24.0,\n  22.0,\n  36.0,\n  34.0],\n [9.0,\n  401.0448852684697,\n  99.02019995940222,\n  100.1249219725039,\n  59.0,\n  43.0,\n  82.0,\n  73.0,\n  168.0,\n  48.0,\n  26.0,\n  20.0,\n  34.0,\n  33.0],\n [10.0,\n  432.0937398296809,\n  101.079176886241,\n  110.0181803158005,\n  60.0,\n  55.0,\n  81.0,\n  77.0,\n  173.0,\n  47.0,\n  30.0,\n  24.0,\n  42.0,\n  36.0],\n [11.0,\n  413.0108957400519,\n  111.0405331399305,\n  105.0190458916858,\n  60.0,\n  55.0,\n  90.0,\n  83.0,\n  161.0,\n  42.0,\n  27.0,\n  23.0,\n  34.0,\n  32.0],\n [12.0,\n  408.0600446012817,\n  102.0,\n  124.0,\n  59.0,\n  46.0,\n  90.0,\n  87.0,\n  179.0,\n  48.0,\n  28.0,\n  27.0,\n  38.0,\n  36.0],\n [13.0,\n  437.1144015014834,\n  102.0441081101697,\n  126.0,\n  68.0,\n  56.0,\n  91.0,\n  87.0,\n  163.0,\n  40.0,\n  30.0,\n  25.0,\n  38.0,\n  32.0],\n [14.0,\n  431.0568407994472,\n  106.0188662455886,\n  118.0169479354554,\n  64.0,\n  47.0,\n  90.0,\n  77.0,\n  177.0,\n  44.0,\n  28.0,\n  27.0,\n  44.0,\n  35.0],\n [15.0,\n  451.0709478563211,\n  115.0173899895142,\n  124.01612798342,\n  65.0,\n  60.0,\n  97.0,\n  84.0,\n  185.0,\n  47.0,\n  30.0,\n  26.0,\n  45.0,\n  35.0],\n [16.0,\n  384.2551235832777,\n  88.09086218218096,\n  105.0190458916858,\n  56.0,\n  49.0,\n  77.0,\n  70.0,\n  170.0,\n  46.0,\n  29.0,\n  25.0,\n  37.0,\n  32.0],\n [17.0,\n  387.0322983938162,\n  90.19977827023745,\n  121.0371843691021,\n  55.0,\n  49.0,\n  84.0,\n  75.0,\n  172.0,\n  46.0,\n  27.0,\n  20.0,\n  38.0,\n  32.0],\n [18.0,\n  400.0799920015996,\n  95.13148795220224,\n  103.0194156457898,\n  61.0,\n  53.0,\n  87.0,\n  80.0,\n  183.0,\n  47.0,\n  28.0,\n  26.0,\n  38.0,\n  34.0],\n [19.0,\n  406.2080747597221,\n  97.51410154434076,\n  99.40824915468535,\n  60.0,\n  47.0,\n  85.0,\n  74.0,\n  177.0,\n  52.0,\n  28.0,\n  28.0,\n  42.0,\n  37.0],\n [20.0,\n  418.01076541161,\n  116.0,\n  146.0136979875518,\n  66.0,\n  50.0,\n  89.0,\n  76.0,\n  187.0,\n  46.0,\n  32.0,\n  26.0,\n  45.0,\n  37.0],\n [21.0,\n  429.0571057563317,\n  116.0689450283753,\n  129.0155029444136,\n  69.0,\n  58.0,\n  95.0,\n  82.0,\n  187.0,\n  46.0,\n  30.0,\n  24.0,\n  39.0,\n  35.0],\n [22.0,\n  445.1898471438898,\n  122.065556157337,\n  124.01612798342,\n  69.0,\n  56.0,\n  96.0,\n  91.0,\n  192.0,\n  51.0,\n  30.0,\n  28.0,\n  47.0,\n  39.0],\n [23.0,\n  449.0,\n  106.0,\n  111.0180165558726,\n  68.0,\n  54.0,\n  95.0,\n  87.0,\n  162.0,\n  40.0,\n  28.0,\n  25.0,\n  36.0,\n  32.0],\n [24.0,\n  379.189926026523,\n  100.0199980003999,\n  108.0,\n  58.0,\n  52.0,\n  82.0,\n  72.0,\n  164.0,\n  42.0,\n  29.0,\n  24.0,\n  43.0,\n  34.0],\n [25.0,\n  415.0433712276345,\n  105.0761628534274,\n  112.0178557195236,\n  62.0,\n  46.0,\n  84.0,\n  78.0,\n  167.0,\n  45.0,\n  29.0,\n  26.0,\n  40.0,\n  36.0],\n [26.0,\n  420.0190471871484,\n  102.0196059588548,\n  100.0199980003999,\n  60.0,\n  55.0,\n  84.0,\n  81.0,\n  173.0,\n  45.0,\n  30.0,\n  24.0,\n  40.0,\n  35.0],\n [27.0,\n  439.1639784863963,\n  111.0720486891279,\n  112.0,\n  59.0,\n  53.0,\n  101.0,\n  85.0,\n  175.0,\n  54.0,\n  30.0,\n  24.0,\n  42.0,\n  32.0],\n [28.0,\n  443.0722288747062,\n  111.0,\n  112.0,\n  65.0,\n  52.0,\n  91.0,\n  85.0,\n  192.0,\n  53.0,\n  32.0,\n  28.0,\n  44.0,\n  37.0],\n [29.0,\n  406.0049260784899,\n  116.1077086157504,\n  115.1086443322134,\n  55.0,\n  36.0,\n  84.0,\n  79.0,\n  195.0,\n  56.0,\n  32.0,\n  27.0,\n  44.0,\n  40.0],\n [30.0,\n  465.0096773186554,\n  103.0436800585072,\n  109.1146186356347,\n  66.0,\n  57.0,\n  98.0,\n  92.0,\n  166.0,\n  46.0,\n  27.0,\n  25.0,\n  37.0,\n  33.0],\n [31.0,\n  469.0095947845844,\n  128.0156240464421,\n  139.0,\n  68.0,\n  61.0,\n  95.0,\n  90.0,\n  176.0,\n  43.0,\n  29.0,\n  23.0,\n  39.0,\n  35.0],\n [32.0,\n  396.1527483181204,\n  97.082439194738,\n  128.0,\n  59.0,\n  55.0,\n  87.0,\n  82.0,\n  182.0,\n  47.0,\n  28.0,\n  26.0,\n  39.0,\n  35.0],\n [33.0,\n  424.0577790820491,\n  97.04638066409278,\n  130.0,\n  60.0,\n  53.0,\n  91.0,\n  85.0,\n  184.0,\n  48.0,\n  28.0,\n  26.0,\n  47.0,\n  37.0],\n [34.0,\n  403.0793966453755,\n  110.0409014866745,\n  117.0,\n  71.0,\n  57.0,\n  94.0,\n  87.0,\n  185.0,\n  44.0,\n  31.0,\n  25.0,\n  42.0,\n  38.0],\n [35.0,\n  408.2756421830722,\n  102.1224754889931,\n  111.0180165558726,\n  57.0,\n  46.0,\n  85.0,\n  83.0,\n  175.0,\n  46.0,\n  26.0,\n  24.0,\n  40.0,\n  35.0],\n [36.0,\n  407.0196555450363,\n  94.04786015641186,\n  135.0148140020198,\n  65.0,\n  54.0,\n  89.0,\n  84.0,\n  192.0,\n  40.0,\n  31.0,\n  28.0,\n  46.0,\n  40.0],\n [37.0,\n  413.0,\n  93.02150289046077,\n  124.2577965360725,\n  60.0,\n  55.0,\n  87.0,\n  80.0,\n  197.0,\n  50.0,\n  30.0,\n  29.0,\n  42.0,\n  42.0],\n [38.0,\n  403.10048375064,\n  116.0689450283753,\n  136.0,\n  67.0,\n  57.0,\n  93.0,\n  87.0,\n  178.0,\n  44.0,\n  27.0,\n  26.0,\n  40.0,\n  37.0],\n [39.0,\n  442.0407221060069,\n  101.0,\n  98.02040603874276,\n  66.0,\n  57.0,\n  93.0,\n  82.0,\n  183.0,\n  43.0,\n  27.0,\n  27.0,\n  42.0,\n  38.0],\n [40.0,\n  433.0103924849842,\n  114.0175425099138,\n  125.015998976131,\n  63.0,\n  52.0,\n  96.0,\n  86.0,\n  177.0,\n  44.0,\n  27.0,\n  24.0,\n  40.0,\n  37.0],\n [41.0,\n  434.028800887683,\n  109.0183470797462,\n  121.0,\n  60.0,\n  55.0,\n  87.0,\n  86.0,\n  178.0,\n  44.0,\n  30.0,\n  26.0,\n  41.0,\n  36.0],\n [42.0,\n  438.0559325017754,\n  98.0,\n  124.0,\n  62.0,\n  54.0,\n  97.0,\n  84.0,\n  182.0,\n  44.0,\n  26.0,\n  26.0,\n  41.0,\n  36.0],\n [43.0,\n  456.0394719758368,\n  109.0183470797462,\n  106.0188662455886,\n  73.0,\n  61.0,\n  95.0,\n  89.0,\n  188.0,\n  48.0,\n  32.0,\n  27.0,\n  43.0,\n  39.0],\n [44.0,\n  395.0455669919611,\n  100.1249219725039,\n  105.0,\n  60.0,\n  54.0,\n  92.0,\n  80.0,\n  190.0,\n  44.0,\n  32.0,\n  28.0,\n  43.0,\n  36.0],\n [45.0,\n  444.4063455892591,\n  108.0416586322146,\n  130.034610777285,\n  66.0,\n  54.0,\n  93.0,\n  87.0,\n  188.0,\n  47.0,\n  30.0,\n  25.0,\n  44.0,\n  37.0],\n [46.0,\n  413.0980513146969,\n  108.0,\n  121.1032617232088,\n  59.0,\n  47.0,\n  84.0,\n  73.0,\n  184.0,\n  48.0,\n  28.0,\n  27.0,\n  40.0,\n  39.0],\n [47.0,\n  410.2060457867485,\n  113.017697729161,\n  110.0181803158005,\n  61.0,\n  52.0,\n  91.0,\n  86.0,\n  186.0,\n  42.0,\n  29.0,\n  26.0,\n  44.0,\n  38.0],\n [48.0,\n  438.0559325017754,\n  101.079176886241,\n  115.0,\n  59.0,\n  46.0,\n  83.0,\n  72.0,\n  196.0,\n  50.0,\n  34.0,\n  28.0,\n  43.0,\n  41.0],\n [49.0,\n  392.0114794237536,\n  98.08159868191383,\n  126.1427762497718,\n  59.0,\n  50.0,\n  80.0,\n  68.0,\n  180.0,\n  46.0,\n  31.0,\n  25.0,\n  42.0,\n  35.0],\n [50.0,\n  379.1068450977903,\n  89.27485648266257,\n  103.1212878119741,\n  50.0,\n  44.0,\n  78.0,\n  76.0,\n  180.0,\n  48.0,\n  26.0,\n  24.0,\n  38.0,\n  34.0],\n [51.0,\n  380.0118419207486,\n  108.0,\n  122.0163923413571,\n  56.0,\n  43.0,\n  77.0,\n  76.0,\n  182.0,\n  49.0,\n  28.0,\n  25.0,\n  42.0,\n  38.0]]"
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 准备训练数据\n",
    "data = readExcel.read_exceldata('A,B,C,D,E,F,G,H,I,J,K,L,M,N', tablePass='lengths.xlsx', print=True)\n",
    "data"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "outputs": [],
   "source": [
    "train_x = []\n",
    "train_y = []\n",
    "validation_x = []\n",
    "validation_y = []\n",
    "loss_acc = 0\n",
    "lossrate_acc = 0\n",
    "m = len(data)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NVIDIA GeForce RTX 2050\n"
     ]
    },
    {
     "data": {
      "text/plain": "'cpu'"
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "if torch.cuda.is_available():\n",
    "    print(torch.cuda.get_device_name(0))\n",
    "    device=torch.device(f'cuda:{0}')\n",
    "else:\n",
    "    device='cpu'\n",
    "device='cpu'        #在这里由于数据量不大，所以用在与GPU数据传输的时间反而大于CPU上直接训练（数据量不大，但是请求等待总线需要时间）\n",
    "device"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "outputs": [],
   "source": [
    "#输入模型与数据集，验证准确度\n",
    "def val_acc(net,val_x,val_y):   #将所有验证集代入模型验证精度，输出各项平均精度\n",
    "    y=net(val_x)\n",
    "    # print(\"prediction:\",y)\n",
    "    error_rate=abs((y-val_y)/val_y)\n",
    "    # print(type(error_rate))\n",
    "    print(\"各项误差:\\n\",error_rate)\n",
    "    ave_error_rate=torch.mean(error_rate, dim=0)\n",
    "    print(\"平均误差：\",ave_error_rate)\n",
    "    return ave_error_rate"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "数据总数： 51\n",
      "第0折交叉验证\n",
      "global_step:0, loss:13280.795\n",
      "global_step:10000, loss:37.953922\n",
      "global_step:20000, loss:31.660564\n",
      "global_step:30000, loss:30.542223\n",
      "global_step:40000, loss:29.06109\n",
      "global_step:50000, loss:27.155954\n",
      "global_step:60000, loss:25.107876\n",
      "global_step:70000, loss:23.212347\n",
      "global_step:80000, loss:21.688837\n",
      "global_step:90000, loss:20.651028\n",
      "global_step:100000, loss:20.12319\n",
      "global_step:110000, loss:19.967533\n",
      "global_step:120000, loss:16.60161\n",
      "global_step:130000, loss:15.117545\n",
      "global_step:140000, loss:14.737069\n",
      "第0折训练完成\n",
      "各项误差:\n",
      " tensor([[5.7490e-03, 1.3999e-02, 3.9292e-02, 2.7068e-02, 2.1650e-02, 4.9186e-02],\n",
      "        [4.1909e-02, 3.1970e-02, 6.5231e-02, 1.2726e-01, 2.4793e-02, 1.2960e-01],\n",
      "        [1.5427e-01, 9.7252e-02, 1.1345e-01, 1.5636e-01, 2.6658e-01, 1.6111e-01],\n",
      "        [1.9228e-02, 1.6304e-02, 3.9157e-02, 1.4971e-02, 7.4545e-02, 8.8789e-02],\n",
      "        [5.5061e-02, 6.2827e-02, 6.7551e-03, 5.1965e-02, 1.0005e-01, 2.5186e-02],\n",
      "        [5.8439e-02, 2.9554e-02, 2.1896e-02, 8.9794e-02, 4.8432e-02, 4.5105e-02],\n",
      "        [3.0765e-02, 4.3482e-02, 5.9887e-02, 1.0816e-01, 1.4097e-01, 5.4131e-02],\n",
      "        [7.0916e-02, 1.4904e-01, 7.6265e-02, 9.3806e-02, 1.1064e-01, 1.0611e-01],\n",
      "        [3.0198e-02, 7.0405e-02, 1.1574e-02, 6.1352e-03, 3.8297e-02, 2.1930e-02],\n",
      "        [1.1652e-02, 8.3220e-03, 7.8638e-02, 4.0683e-02, 8.8225e-02, 4.0988e-02],\n",
      "        [1.0090e-02, 3.9147e-02, 4.3970e-02, 8.3661e-03, 1.2979e-04, 4.5504e-02]],\n",
      "       grad_fn=<AbsBackward0>)\n",
      "平均误差： tensor([0.0444, 0.0511, 0.0506, 0.0659, 0.0831, 0.0698],\n",
      "       grad_fn=<MeanBackward1>)\n",
      "第1折交叉验证\n",
      "global_step:0, loss:8602.1709\n",
      "global_step:10000, loss:34.806267\n",
      "global_step:20000, loss:29.829704\n",
      "global_step:30000, loss:23.035528\n",
      "global_step:40000, loss:17.354654\n",
      "global_step:50000, loss:14.445484\n",
      "global_step:60000, loss:12.836862\n",
      "global_step:70000, loss:11.802411\n",
      "global_step:80000, loss:10.992797\n",
      "global_step:90000, loss:10.177583\n",
      "global_step:100000, loss:10.090493\n",
      "global_step:110000, loss:10.050189\n",
      "global_step:120000, loss:9.971818\n",
      "global_step:130000, loss:9.9306402\n",
      "global_step:140000, loss:9.8950949\n",
      "第1折训练完成\n",
      "各项误差:\n",
      " tensor([[0.0433, 0.0123, 0.0887, 0.0287, 0.0718, 0.0935],\n",
      "        [0.1141, 0.1739, 0.0169, 0.1159, 0.0985, 0.1227],\n",
      "        [0.0367, 0.0009, 0.0240, 0.0576, 0.1335, 0.0247],\n",
      "        [0.0588, 0.0087, 0.0558, 0.2552, 0.1185, 0.1208],\n",
      "        [0.0899, 0.1169, 0.0508, 0.1175, 0.1544, 0.1121],\n",
      "        [0.0529, 0.1080, 0.0143, 0.0811, 0.0192, 0.1275],\n",
      "        [0.0056, 0.0847, 0.0429, 0.0238, 0.0929, 0.0193],\n",
      "        [0.1018, 0.1334, 0.0565, 0.1498, 0.0152, 0.1732],\n",
      "        [0.0006, 0.0230, 0.1216, 0.0202, 0.0222, 0.0121],\n",
      "        [0.0753, 0.0973, 0.0522, 0.0761, 0.1246, 0.1156]],\n",
      "       grad_fn=<AbsBackward0>)\n",
      "平均误差： tensor([0.0579, 0.0759, 0.0524, 0.0926, 0.0851, 0.0921],\n",
      "       grad_fn=<MeanBackward1>)\n",
      "第2折交叉验证\n",
      "global_step:0, loss:3733.1433\n",
      "global_step:10000, loss:29.022358\n",
      "global_step:20000, loss:25.25638\n",
      "global_step:30000, loss:24.254351\n",
      "global_step:40000, loss:22.96895\n",
      "global_step:50000, loss:21.285769\n",
      "global_step:60000, loss:19.303171\n",
      "global_step:70000, loss:17.230892\n",
      "global_step:80000, loss:15.456261\n",
      "global_step:90000, loss:14.335218\n",
      "global_step:100000, loss:13.764626\n",
      "global_step:110000, loss:13.576373\n",
      "global_step:120000, loss:13.515482\n",
      "global_step:130000, loss:12.888103\n",
      "global_step:140000, loss:12.83013\n",
      "第2折训练完成\n",
      "各项误差:\n",
      " tensor([[0.0057, 0.0007, 0.0286, 0.0065, 0.0887, 0.0115],\n",
      "        [0.0666, 0.0538, 0.2041, 0.1273, 0.1278, 0.0254],\n",
      "        [0.1362, 0.1048, 0.0140, 0.0350, 0.1169, 0.1599],\n",
      "        [0.0366, 0.0516, 0.0236, 0.0419, 0.0690, 0.0384],\n",
      "        [0.0970, 0.1221, 0.0365, 0.0030, 0.1399, 0.1089],\n",
      "        [0.0796, 0.1310, 0.1005, 0.1094, 0.0750, 0.0458],\n",
      "        [0.0046, 0.0363, 0.0246, 0.0168, 0.1170, 0.0015],\n",
      "        [0.0832, 0.0094, 0.1325, 0.0272, 0.1030, 0.0420],\n",
      "        [0.0643, 0.1256, 0.0704, 0.0737, 0.0376, 0.1077],\n",
      "        [0.1307, 0.0531, 0.1800, 0.1354, 0.0884, 0.1685]],\n",
      "       grad_fn=<AbsBackward0>)\n",
      "平均误差： tensor([0.0704, 0.0688, 0.0815, 0.0576, 0.0963, 0.0710],\n",
      "       grad_fn=<MeanBackward1>)\n",
      "第3折交叉验证\n",
      "global_step:0, loss:6287.3257\n",
      "global_step:10000, loss:33.353424\n",
      "global_step:20000, loss:27.256475\n",
      "global_step:30000, loss:25.133936\n",
      "global_step:40000, loss:22.544275\n",
      "global_step:50000, loss:20.311684\n",
      "global_step:60000, loss:18.402559\n",
      "global_step:70000, loss:16.86134\n",
      "global_step:80000, loss:14.474017\n",
      "global_step:90000, loss:13.420537\n",
      "global_step:100000, loss:13.242588\n",
      "global_step:110000, loss:11.158134\n",
      "global_step:120000, loss:10.387848\n",
      "global_step:130000, loss:10.145474\n",
      "global_step:140000, loss:9.9769096\n",
      "第3折训练完成\n",
      "各项误差:\n",
      " tensor([[7.0450e-02, 7.2161e-02, 7.5599e-02, 9.7720e-02, 1.2699e-01, 7.4231e-02],\n",
      "        [1.3500e-01, 4.5305e-02, 1.9765e-01, 3.5460e-01, 2.7789e-01, 1.7442e-01],\n",
      "        [2.3361e-03, 3.0798e-02, 4.0390e-02, 5.9685e-02, 7.4645e-02, 3.9563e-02],\n",
      "        [6.0425e-03, 1.1430e-01, 3.8564e-02, 1.1083e-01, 4.7137e-02, 4.7035e-02],\n",
      "        [8.9733e-02, 1.0519e-01, 3.5683e-02, 1.8692e-02, 6.7426e-02, 1.5021e-02],\n",
      "        [1.7270e-01, 1.2117e-01, 9.3396e-02, 1.8949e-01, 1.7565e-01, 1.7933e-01],\n",
      "        [1.0092e-02, 1.4643e-04, 4.7386e-02, 3.7560e-02, 4.5976e-03, 3.1260e-02],\n",
      "        [7.5152e-02, 2.3145e-01, 2.0055e-01, 3.0721e-02, 3.9850e-02, 1.7574e-03],\n",
      "        [3.0406e-02, 1.0691e-01, 1.0521e-01, 1.0903e-01, 4.7241e-02, 2.4757e-02],\n",
      "        [1.3051e-02, 4.1986e-02, 9.8512e-02, 1.1688e-02, 3.5850e-02, 2.0603e-03]],\n",
      "       grad_fn=<AbsBackward0>)\n",
      "平均误差： tensor([0.0605, 0.0869, 0.0933, 0.1020, 0.0897, 0.0589],\n",
      "       grad_fn=<MeanBackward1>)\n",
      "第4折交叉验证\n",
      "global_step:0, loss:4583.8457\n",
      "global_step:10000, loss:35.366711\n",
      "global_step:20000, loss:30.807997\n",
      "global_step:30000, loss:25.442392\n",
      "global_step:40000, loss:20.177736\n",
      "global_step:50000, loss:17.796194\n",
      "global_step:60000, loss:16.63019\n",
      "global_step:70000, loss:15.844965\n",
      "global_step:80000, loss:14.915243\n",
      "global_step:90000, loss:14.106473\n",
      "global_step:100000, loss:13.67067\n",
      "global_step:110000, loss:13.369753\n",
      "global_step:120000, loss:12.592485\n",
      "global_step:130000, loss:12.21694\n",
      "global_step:140000, loss:12.107687\n",
      "第4折训练完成\n",
      "各项误差:\n",
      " tensor([[0.0631, 0.0296, 0.0516, 0.0383, 0.0239, 0.0667],\n",
      "        [0.0215, 0.0569, 0.0752, 0.0095, 0.0730, 0.0469],\n",
      "        [0.0400, 0.0343, 0.0404, 0.0292, 0.1015, 0.0107],\n",
      "        [0.0373, 0.0450, 0.0885, 0.0091, 0.0589, 0.0179],\n",
      "        [0.0467, 0.0467, 0.0153, 0.0469, 0.0070, 0.0257],\n",
      "        [0.1149, 0.0198, 0.1121, 0.0494, 0.1352, 0.1181],\n",
      "        [0.0230, 0.0101, 0.0752, 0.0112, 0.0283, 0.0211],\n",
      "        [0.0361, 0.0130, 0.0394, 0.0183, 0.0239, 0.0708],\n",
      "        [0.0610, 0.0601, 0.0435, 0.0091, 0.0766, 0.0358],\n",
      "        [0.0712, 0.0488, 0.0367, 0.0156, 0.0075, 0.0116]],\n",
      "       grad_fn=<AbsBackward0>)\n",
      "平均误差： tensor([0.0515, 0.0364, 0.0578, 0.0237, 0.0536, 0.0425],\n",
      "       grad_fn=<MeanBackward1>)\n",
      "time cost 175.66089797019958 s\n"
     ]
    }
   ],
   "source": [
    "print(\"数据总数：\",str(m))\n",
    "time_start = time.time()\n",
    "\n",
    "trainloss_best = []      #记录每一折的训练loss\n",
    "trainloss_last = []     #清空\n",
    "val_acc_all=[]\n",
    "for k in range(k_cross):\n",
    "    print(\"第%d折交叉验证\" % k)\n",
    "    train_x = []    #清空\n",
    "    train_y = []\n",
    "    validation_x = []\n",
    "    validation_y = []\n",
    "    for i in range(m):\n",
    "        if i % k_cross == k:  # 每5张中有一张放入验证集\n",
    "            validation_x.append([data[i][1], data[i][2], data[i][3], data[i][4], data[i][5], data[i][6], data[i][7]])\n",
    "            validation_y.append([data[i][8], data[i][9], data[i][10], data[i][11], data[i][12], data[i][13]])\n",
    "            # print(\"已读入验证集\" + str(data[i][0]))\n",
    "        else:\n",
    "            train_x.append([data[i][1], data[i][2], data[i][3], data[i][4], data[i][5], data[i][6], data[i][7]])\n",
    "            train_y.append([data[i][8], data[i][9], data[i][10], data[i][11], data[i][12], data[i][13]])\n",
    "            # print(\"已读入训练集\" + str(data[i][0]))\n",
    "\n",
    "    # 张量化\n",
    "    x = torch.tensor(train_x).to(torch.float32).to(device)\n",
    "    y = torch.tensor(train_y).to(torch.float32).to(device)\n",
    "    vx = torch.tensor(validation_x).to(torch.float32).to(device)\n",
    "    vy = torch.tensor(validation_y).to(torch.float32).to(device)\n",
    "\n",
    "    if trainModel:      #从k折交叉验证数据集重新训练\n",
    "        net = Net(net_frame).to(device)  # 生成模型对象\n",
    "        # print(net)\n",
    "\n",
    "        # 定义训练方法,损失函数,日志间隔\n",
    "        optimizer = torch.optim.Adam(net.parameters(), lr=learning_rate)  # .Adam:adam优化器 ,lr:learning_rate 学习率\n",
    "        loss_func = torch.nn.MSELoss()  # MSELoss损失函数\n",
    "        log_step_interval = 10000  # 每100轮训练记录一次loss日志用于后期绘图\n",
    "\n",
    "        # 开始训练\n",
    "        loss_best = 10000\n",
    "        lose_last = 0\n",
    "        best_save = False  # 保存最优模型\n",
    "        for t in range(iteration):\n",
    "            # 前向计算->计算损失函数->(从损失函数)反向传播->更新网络\n",
    "            prediction = net(x)\n",
    "            loss = loss_func(prediction, y)\n",
    "            optimizer.zero_grad()  # 清空梯度（可以不写）\n",
    "            loss.backward()  # 反向传播计算梯度\n",
    "            optimizer.step()  # 更新网络\n",
    "\n",
    "            if t % log_step_interval == 0:\n",
    "                # 控制台输出一下\n",
    "                # print(str(t)+\"Loss = %.8f\" % loss.data)\n",
    "                global_iter_num = t\n",
    "                print(\"global_step:{}, loss:{:.8}\".format(global_iter_num, loss.item()))\n",
    "                lose_last = loss.item()\n",
    "                if lose_last < loss_best:\n",
    "                    loss_best = lose_last\n",
    "        print(\"第%d折训练完成\"%k)\n",
    "        trainloss_best.append(loss_best)\n",
    "        trainloss_last.append(lose_last)\n",
    "        #验证集合验证\n",
    "        val_acc_all.append(val_acc(net,vx,vy))\n",
    "\n",
    "    else:       #直接调用已训练好模型重新评估\n",
    "        net = torch.load('net/last.pt')\n",
    "        pass\n",
    "time_end = time.time()\n",
    "print('time cost', time_end - time_start, 's')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "trainloss_best:[14.737069129943848, 9.895094871520996, 12.830129623413086, 9.976909637451172, 12.107686996459961]\n",
      "trainloss_best:[14.737069129943848, 9.895094871520996, 12.830129623413086, 9.976909637451172, 12.107686996459961]\n",
      "val_acc_all:[tensor([0.0444, 0.0511, 0.0506, 0.0659, 0.0831, 0.0698],\n",
      "       grad_fn=<MeanBackward1>), tensor([0.0579, 0.0759, 0.0524, 0.0926, 0.0851, 0.0921],\n",
      "       grad_fn=<MeanBackward1>), tensor([0.0704, 0.0688, 0.0815, 0.0576, 0.0963, 0.0710],\n",
      "       grad_fn=<MeanBackward1>), tensor([0.0605, 0.0869, 0.0933, 0.1020, 0.0897, 0.0589],\n",
      "       grad_fn=<MeanBackward1>), tensor([0.0515, 0.0364, 0.0578, 0.0237, 0.0536, 0.0425],\n",
      "       grad_fn=<MeanBackward1>)]\n"
     ]
    }
   ],
   "source": [
    "print(\"trainloss_best:\"+str(trainloss_best))\n",
    "print(\"trainloss_best:\"+str(trainloss_last))\n",
    "print(\"val_acc_all:\"+str(val_acc_all))"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['目测身高', '目测肩宽', '目测大臂长', '目测小臂长', '目测大腿长', '目测小腿长']\n",
      "[0.044388823211193085, 0.051118072122335434, 0.0505562424659729, 0.06586953997612, 0.0831184908747673, 0.06978590041399002]\n",
      "[0.05790172144770622, 0.07589936256408691, 0.052368033677339554, 0.09259636700153351, 0.08508194237947464, 0.09214834868907928]\n",
      "[0.07044940441846848, 0.06881637871265411, 0.08147630840539932, 0.057625912129879, 0.09634065628051758, 0.07096338272094727]\n",
      "[0.06049622967839241, 0.08694088459014893, 0.09329331666231155, 0.10200075060129166, 0.08972788602113724, 0.05894298478960991]\n",
      "[0.05147934705018997, 0.036426037549972534, 0.05779798701405525, 0.02367136999964714, 0.0535852424800396, 0.0425269640982151]\n"
     ]
    }
   ],
   "source": [
    "lable=['目测身高','目测肩宽','目测大臂长','目测小臂长','目测大腿长','目测小腿长']\n",
    "print(lable)\n",
    "val_acc_all_list=[]\n",
    "for e in val_acc_all:\n",
    "    val_acc_all_list.append(e.tolist())\n",
    "for e in val_acc_all_list:\n",
    "    print(e)\n"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['目测身高', '目测肩宽', '目测大臂长', '目测小臂长', '目测大腿长', '目测小腿长'] 总平均误差\n",
      "[0.05694310739636421, 0.06384014338254929, 0.06709837913513184, 0.06835278123617172, 0.08157084137201309, 0.06687351316213608]\n"
     ]
    }
   ],
   "source": [
    "lable=['目测身高','目测肩宽','目测大臂长','目测小臂长','目测大腿长','目测小腿长']\n",
    "print(lable,'总平均误差')\n",
    "print(torch.mean(torch.tensor(val_acc_all_list),dim=0).tolist())"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}