learn_pytorch_24_pytorch实战项目之多元线性回归函数拟合.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 本例说明：拟合一个多项式函数 "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 1 导入需要的模块\n",
    "import torch\n",
    "import numpy as np \n",
    "import matplotlib.pyplot as plt \n",
    "import torch.nn as nn "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "## 2 生成数据集  \n",
    "x_ = np.arange(-5, 5, 0.1, dtype=np.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-5.0000000e+00, -4.9000001e+00, -4.8000002e+00, -4.7000003e+00,\n",
       "       -4.6000004e+00, -4.5000005e+00, -4.4000006e+00, -4.3000007e+00,\n",
       "       -4.2000008e+00, -4.1000009e+00, -4.0000010e+00, -3.9000010e+00,\n",
       "       -3.8000011e+00, -3.7000012e+00, -3.6000013e+00, -3.5000014e+00,\n",
       "       -3.4000015e+00, -3.3000016e+00, -3.2000017e+00, -3.1000018e+00,\n",
       "       -3.0000019e+00, -2.9000020e+00, -2.8000021e+00, -2.7000022e+00,\n",
       "       -2.6000023e+00, -2.5000024e+00, -2.4000025e+00, -2.3000026e+00,\n",
       "       -2.2000027e+00, -2.1000028e+00, -2.0000029e+00, -1.9000030e+00,\n",
       "       -1.8000031e+00, -1.7000031e+00, -1.6000032e+00, -1.5000033e+00,\n",
       "       -1.4000034e+00, -1.3000035e+00, -1.2000036e+00, -1.1000037e+00,\n",
       "       -1.0000038e+00, -9.0000391e-01, -8.0000401e-01, -7.0000410e-01,\n",
       "       -6.0000420e-01, -5.0000429e-01, -4.0000439e-01, -3.0000448e-01,\n",
       "       -2.0000458e-01, -1.0000467e-01, -4.7683716e-06,  9.9995136e-02,\n",
       "        1.9999504e-01,  2.9999495e-01,  3.9999485e-01,  4.9999475e-01,\n",
       "        5.9999466e-01,  6.9999456e-01,  7.9999447e-01,  8.9999437e-01,\n",
       "        9.9999428e-01,  1.0999942e+00,  1.1999941e+00,  1.2999940e+00,\n",
       "        1.3999939e+00,  1.4999938e+00,  1.5999937e+00,  1.6999936e+00,\n",
       "        1.7999935e+00,  1.8999934e+00,  1.9999933e+00,  2.0999932e+00,\n",
       "        2.1999931e+00,  2.2999930e+00,  2.3999929e+00,  2.4999928e+00,\n",
       "        2.5999928e+00,  2.6999927e+00,  2.7999926e+00,  2.8999925e+00,\n",
       "        2.9999924e+00,  3.0999928e+00,  3.1999922e+00,  3.2999916e+00,\n",
       "        3.3999920e+00,  3.4999924e+00,  3.5999918e+00,  3.6999912e+00,\n",
       "        3.7999916e+00,  3.8999920e+00,  3.9999914e+00,  4.0999908e+00,\n",
       "        4.1999912e+00,  4.2999916e+00,  4.3999910e+00,  4.4999905e+00,\n",
       "        4.5999908e+00,  4.6999912e+00,  4.7999907e+00,  4.8999901e+00],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2.1 第一维特征  \n",
    "x_data1 = np.transpose([x_])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-5.0000000e+00],\n",
       "       [-4.9000001e+00],\n",
       "       [-4.8000002e+00],\n",
       "       [-4.7000003e+00],\n",
       "       [-4.6000004e+00],\n",
       "       [-4.5000005e+00],\n",
       "       [-4.4000006e+00],\n",
       "       [-4.3000007e+00],\n",
       "       [-4.2000008e+00],\n",
       "       [-4.1000009e+00],\n",
       "       [-4.0000010e+00],\n",
       "       [-3.9000010e+00],\n",
       "       [-3.8000011e+00],\n",
       "       [-3.7000012e+00],\n",
       "       [-3.6000013e+00],\n",
       "       [-3.5000014e+00],\n",
       "       [-3.4000015e+00],\n",
       "       [-3.3000016e+00],\n",
       "       [-3.2000017e+00],\n",
       "       [-3.1000018e+00],\n",
       "       [-3.0000019e+00],\n",
       "       [-2.9000020e+00],\n",
       "       [-2.8000021e+00],\n",
       "       [-2.7000022e+00],\n",
       "       [-2.6000023e+00],\n",
       "       [-2.5000024e+00],\n",
       "       [-2.4000025e+00],\n",
       "       [-2.3000026e+00],\n",
       "       [-2.2000027e+00],\n",
       "       [-2.1000028e+00],\n",
       "       [-2.0000029e+00],\n",
       "       [-1.9000030e+00],\n",
       "       [-1.8000031e+00],\n",
       "       [-1.7000031e+00],\n",
       "       [-1.6000032e+00],\n",
       "       [-1.5000033e+00],\n",
       "       [-1.4000034e+00],\n",
       "       [-1.3000035e+00],\n",
       "       [-1.2000036e+00],\n",
       "       [-1.1000037e+00],\n",
       "       [-1.0000038e+00],\n",
       "       [-9.0000391e-01],\n",
       "       [-8.0000401e-01],\n",
       "       [-7.0000410e-01],\n",
       "       [-6.0000420e-01],\n",
       "       [-5.0000429e-01],\n",
       "       [-4.0000439e-01],\n",
       "       [-3.0000448e-01],\n",
       "       [-2.0000458e-01],\n",
       "       [-1.0000467e-01],\n",
       "       [-4.7683716e-06],\n",
       "       [ 9.9995136e-02],\n",
       "       [ 1.9999504e-01],\n",
       "       [ 2.9999495e-01],\n",
       "       [ 3.9999485e-01],\n",
       "       [ 4.9999475e-01],\n",
       "       [ 5.9999466e-01],\n",
       "       [ 6.9999456e-01],\n",
       "       [ 7.9999447e-01],\n",
       "       [ 8.9999437e-01],\n",
       "       [ 9.9999428e-01],\n",
       "       [ 1.0999942e+00],\n",
       "       [ 1.1999941e+00],\n",
       "       [ 1.2999940e+00],\n",
       "       [ 1.3999939e+00],\n",
       "       [ 1.4999938e+00],\n",
       "       [ 1.5999937e+00],\n",
       "       [ 1.6999936e+00],\n",
       "       [ 1.7999935e+00],\n",
       "       [ 1.8999934e+00],\n",
       "       [ 1.9999933e+00],\n",
       "       [ 2.0999932e+00],\n",
       "       [ 2.1999931e+00],\n",
       "       [ 2.2999930e+00],\n",
       "       [ 2.3999929e+00],\n",
       "       [ 2.4999928e+00],\n",
       "       [ 2.5999928e+00],\n",
       "       [ 2.6999927e+00],\n",
       "       [ 2.7999926e+00],\n",
       "       [ 2.8999925e+00],\n",
       "       [ 2.9999924e+00],\n",
       "       [ 3.0999928e+00],\n",
       "       [ 3.1999922e+00],\n",
       "       [ 3.2999916e+00],\n",
       "       [ 3.3999920e+00],\n",
       "       [ 3.4999924e+00],\n",
       "       [ 3.5999918e+00],\n",
       "       [ 3.6999912e+00],\n",
       "       [ 3.7999916e+00],\n",
       "       [ 3.8999920e+00],\n",
       "       [ 3.9999914e+00],\n",
       "       [ 4.0999908e+00],\n",
       "       [ 4.1999912e+00],\n",
       "       [ 4.2999916e+00],\n",
       "       [ 4.3999910e+00],\n",
       "       [ 4.4999905e+00],\n",
       "       [ 4.5999908e+00],\n",
       "       [ 4.6999912e+00],\n",
       "       [ 4.7999907e+00],\n",
       "       [ 4.8999901e+00]], dtype=float32)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_data1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#  2.2 第二维特征\n",
    "x_data2 = np.power(x_data1, 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[2.5000000e+01],\n",
       "       [2.4010000e+01],\n",
       "       [2.3040001e+01],\n",
       "       [2.2090002e+01],\n",
       "       [2.1160004e+01],\n",
       "       [2.0250004e+01],\n",
       "       [1.9360004e+01],\n",
       "       [1.8490005e+01],\n",
       "       [1.7640007e+01],\n",
       "       [1.6810007e+01],\n",
       "       [1.6000008e+01],\n",
       "       [1.5210009e+01],\n",
       "       [1.4440009e+01],\n",
       "       [1.3690009e+01],\n",
       "       [1.2960010e+01],\n",
       "       [1.2250010e+01],\n",
       "       [1.1560010e+01],\n",
       "       [1.0890011e+01],\n",
       "       [1.0240011e+01],\n",
       "       [9.6100111e+00],\n",
       "       [9.0000114e+00],\n",
       "       [8.4100113e+00],\n",
       "       [7.8400116e+00],\n",
       "       [7.2900119e+00],\n",
       "       [6.7600117e+00],\n",
       "       [6.2500119e+00],\n",
       "       [5.7600117e+00],\n",
       "       [5.2900119e+00],\n",
       "       [4.8400116e+00],\n",
       "       [4.4100118e+00],\n",
       "       [4.0000114e+00],\n",
       "       [3.6100113e+00],\n",
       "       [3.2400110e+00],\n",
       "       [2.8900106e+00],\n",
       "       [2.5600104e+00],\n",
       "       [2.2500100e+00],\n",
       "       [1.9600096e+00],\n",
       "       [1.6900091e+00],\n",
       "       [1.4400086e+00],\n",
       "       [1.2100081e+00],\n",
       "       [1.0000076e+00],\n",
       "       [8.1000704e-01],\n",
       "       [6.4000642e-01],\n",
       "       [4.9000573e-01],\n",
       "       [3.6000502e-01],\n",
       "       [2.5000429e-01],\n",
       "       [1.6000351e-01],\n",
       "       [9.0002686e-02],\n",
       "       [4.0001832e-02],\n",
       "       [1.0000935e-02],\n",
       "       [2.2737368e-11],\n",
       "       [9.9990275e-03],\n",
       "       [3.9998017e-02],\n",
       "       [8.9996964e-02],\n",
       "       [1.5999588e-01],\n",
       "       [2.4999475e-01],\n",
       "       [3.5999358e-01],\n",
       "       [4.8999238e-01],\n",
       "       [6.3999116e-01],\n",
       "       [8.0998987e-01],\n",
       "       [9.9998856e-01],\n",
       "       [1.2099872e+00],\n",
       "       [1.4399858e+00],\n",
       "       [1.6899843e+00],\n",
       "       [1.9599829e+00],\n",
       "       [2.2499814e+00],\n",
       "       [2.5599799e+00],\n",
       "       [2.8899782e+00],\n",
       "       [3.2399766e+00],\n",
       "       [3.6099751e+00],\n",
       "       [3.9999733e+00],\n",
       "       [4.4099717e+00],\n",
       "       [4.8399696e+00],\n",
       "       [5.2899680e+00],\n",
       "       [5.7599659e+00],\n",
       "       [6.2499642e+00],\n",
       "       [6.7599621e+00],\n",
       "       [7.2899604e+00],\n",
       "       [7.8399582e+00],\n",
       "       [8.4099560e+00],\n",
       "       [8.9999542e+00],\n",
       "       [9.6099548e+00],\n",
       "       [1.0239950e+01],\n",
       "       [1.0889945e+01],\n",
       "       [1.1559945e+01],\n",
       "       [1.2249947e+01],\n",
       "       [1.2959941e+01],\n",
       "       [1.3689935e+01],\n",
       "       [1.4439937e+01],\n",
       "       [1.5209937e+01],\n",
       "       [1.5999931e+01],\n",
       "       [1.6809925e+01],\n",
       "       [1.7639927e+01],\n",
       "       [1.8489927e+01],\n",
       "       [1.9359921e+01],\n",
       "       [2.0249914e+01],\n",
       "       [2.1159916e+01],\n",
       "       [2.2089918e+01],\n",
       "       [2.3039909e+01],\n",
       "       [2.4009903e+01]], dtype=float32)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_data2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2.3 合并特征 \n",
    "x_data = np.hstack((x_data1, x_data2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-5.0000000e+00,  2.5000000e+01],\n",
       "       [-4.9000001e+00,  2.4010000e+01],\n",
       "       [-4.8000002e+00,  2.3040001e+01],\n",
       "       [-4.7000003e+00,  2.2090002e+01],\n",
       "       [-4.6000004e+00,  2.1160004e+01],\n",
       "       [-4.5000005e+00,  2.0250004e+01],\n",
       "       [-4.4000006e+00,  1.9360004e+01],\n",
       "       [-4.3000007e+00,  1.8490005e+01],\n",
       "       [-4.2000008e+00,  1.7640007e+01],\n",
       "       [-4.1000009e+00,  1.6810007e+01],\n",
       "       [-4.0000010e+00,  1.6000008e+01],\n",
       "       [-3.9000010e+00,  1.5210009e+01],\n",
       "       [-3.8000011e+00,  1.4440009e+01],\n",
       "       [-3.7000012e+00,  1.3690009e+01],\n",
       "       [-3.6000013e+00,  1.2960010e+01],\n",
       "       [-3.5000014e+00,  1.2250010e+01],\n",
       "       [-3.4000015e+00,  1.1560010e+01],\n",
       "       [-3.3000016e+00,  1.0890011e+01],\n",
       "       [-3.2000017e+00,  1.0240011e+01],\n",
       "       [-3.1000018e+00,  9.6100111e+00],\n",
       "       [-3.0000019e+00,  9.0000114e+00],\n",
       "       [-2.9000020e+00,  8.4100113e+00],\n",
       "       [-2.8000021e+00,  7.8400116e+00],\n",
       "       [-2.7000022e+00,  7.2900119e+00],\n",
       "       [-2.6000023e+00,  6.7600117e+00],\n",
       "       [-2.5000024e+00,  6.2500119e+00],\n",
       "       [-2.4000025e+00,  5.7600117e+00],\n",
       "       [-2.3000026e+00,  5.2900119e+00],\n",
       "       [-2.2000027e+00,  4.8400116e+00],\n",
       "       [-2.1000028e+00,  4.4100118e+00],\n",
       "       [-2.0000029e+00,  4.0000114e+00],\n",
       "       [-1.9000030e+00,  3.6100113e+00],\n",
       "       [-1.8000031e+00,  3.2400110e+00],\n",
       "       [-1.7000031e+00,  2.8900106e+00],\n",
       "       [-1.6000032e+00,  2.5600104e+00],\n",
       "       [-1.5000033e+00,  2.2500100e+00],\n",
       "       [-1.4000034e+00,  1.9600096e+00],\n",
       "       [-1.3000035e+00,  1.6900091e+00],\n",
       "       [-1.2000036e+00,  1.4400086e+00],\n",
       "       [-1.1000037e+00,  1.2100081e+00],\n",
       "       [-1.0000038e+00,  1.0000076e+00],\n",
       "       [-9.0000391e-01,  8.1000704e-01],\n",
       "       [-8.0000401e-01,  6.4000642e-01],\n",
       "       [-7.0000410e-01,  4.9000573e-01],\n",
       "       [-6.0000420e-01,  3.6000502e-01],\n",
       "       [-5.0000429e-01,  2.5000429e-01],\n",
       "       [-4.0000439e-01,  1.6000351e-01],\n",
       "       [-3.0000448e-01,  9.0002686e-02],\n",
       "       [-2.0000458e-01,  4.0001832e-02],\n",
       "       [-1.0000467e-01,  1.0000935e-02],\n",
       "       [-4.7683716e-06,  2.2737368e-11],\n",
       "       [ 9.9995136e-02,  9.9990275e-03],\n",
       "       [ 1.9999504e-01,  3.9998017e-02],\n",
       "       [ 2.9999495e-01,  8.9996964e-02],\n",
       "       [ 3.9999485e-01,  1.5999588e-01],\n",
       "       [ 4.9999475e-01,  2.4999475e-01],\n",
       "       [ 5.9999466e-01,  3.5999358e-01],\n",
       "       [ 6.9999456e-01,  4.8999238e-01],\n",
       "       [ 7.9999447e-01,  6.3999116e-01],\n",
       "       [ 8.9999437e-01,  8.0998987e-01],\n",
       "       [ 9.9999428e-01,  9.9998856e-01],\n",
       "       [ 1.0999942e+00,  1.2099872e+00],\n",
       "       [ 1.1999941e+00,  1.4399858e+00],\n",
       "       [ 1.2999940e+00,  1.6899843e+00],\n",
       "       [ 1.3999939e+00,  1.9599829e+00],\n",
       "       [ 1.4999938e+00,  2.2499814e+00],\n",
       "       [ 1.5999937e+00,  2.5599799e+00],\n",
       "       [ 1.6999936e+00,  2.8899782e+00],\n",
       "       [ 1.7999935e+00,  3.2399766e+00],\n",
       "       [ 1.8999934e+00,  3.6099751e+00],\n",
       "       [ 1.9999933e+00,  3.9999733e+00],\n",
       "       [ 2.0999932e+00,  4.4099717e+00],\n",
       "       [ 2.1999931e+00,  4.8399696e+00],\n",
       "       [ 2.2999930e+00,  5.2899680e+00],\n",
       "       [ 2.3999929e+00,  5.7599659e+00],\n",
       "       [ 2.4999928e+00,  6.2499642e+00],\n",
       "       [ 2.5999928e+00,  6.7599621e+00],\n",
       "       [ 2.6999927e+00,  7.2899604e+00],\n",
       "       [ 2.7999926e+00,  7.8399582e+00],\n",
       "       [ 2.8999925e+00,  8.4099560e+00],\n",
       "       [ 2.9999924e+00,  8.9999542e+00],\n",
       "       [ 3.0999928e+00,  9.6099548e+00],\n",
       "       [ 3.1999922e+00,  1.0239950e+01],\n",
       "       [ 3.2999916e+00,  1.0889945e+01],\n",
       "       [ 3.3999920e+00,  1.1559945e+01],\n",
       "       [ 3.4999924e+00,  1.2249947e+01],\n",
       "       [ 3.5999918e+00,  1.2959941e+01],\n",
       "       [ 3.6999912e+00,  1.3689935e+01],\n",
       "       [ 3.7999916e+00,  1.4439937e+01],\n",
       "       [ 3.8999920e+00,  1.5209937e+01],\n",
       "       [ 3.9999914e+00,  1.5999931e+01],\n",
       "       [ 4.0999908e+00,  1.6809925e+01],\n",
       "       [ 4.1999912e+00,  1.7639927e+01],\n",
       "       [ 4.2999916e+00,  1.8489927e+01],\n",
       "       [ 4.3999910e+00,  1.9359921e+01],\n",
       "       [ 4.4999905e+00,  2.0249914e+01],\n",
       "       [ 4.5999908e+00,  2.1159916e+01],\n",
       "       [ 4.6999912e+00,  2.2089918e+01],\n",
       "       [ 4.7999907e+00,  2.3039909e+01],\n",
       "       [ 4.8999901e+00,  2.4009903e+01]], dtype=float32)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2.4 目标值：设置系数1.5和2\n",
    "noise = np.random.randn(100)  # 添加噪声\n",
    "y_data = 1.5 * x_data[:, 0] + 2 * x_data[:, 1]+noise"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 4.11419670e+01,  4.07796440e+01,  3.82623782e+01,  3.53068261e+01,\n",
       "        3.58943766e+01,  3.36611491e+01,  3.09941252e+01,  2.88306911e+01,\n",
       "        2.73769809e+01,  2.72507560e+01,  2.60401126e+01,  2.55601984e+01,\n",
       "        2.41441802e+01,  2.26711588e+01,  1.99659162e+01,  1.93178349e+01,\n",
       "        1.65910443e+01,  1.62756602e+01,  1.62694147e+01,  1.41186503e+01,\n",
       "        1.27223597e+01,  1.30499395e+01,  9.73653235e+00,  1.14849502e+01,\n",
       "        1.08780236e+01,  8.66170703e+00,  9.31852056e+00,  7.41446453e+00,\n",
       "        6.59124377e+00,  3.35320460e+00,  5.21888775e+00,  4.77551093e+00,\n",
       "        1.54413105e+00,  2.00182115e+00,  4.44254979e+00,  2.30804944e+00,\n",
       "        2.51835108e+00,  2.00093308e+00,  8.03879993e-01, -6.21852906e-01,\n",
       "       -5.98419610e-02,  1.19503541e-01,  2.54667937e-01,  2.91466927e-01,\n",
       "        5.25959703e-01,  1.80523510e+00, -8.96105350e-01,  1.36584010e+00,\n",
       "        2.92602653e-03,  5.13342386e-01,  7.83195804e-03, -7.40093247e-01,\n",
       "        7.16789514e-01,  1.54625082e+00,  3.20021642e+00,  1.23638592e+00,\n",
       "       -1.10464369e+00,  2.31137383e+00,  2.68551716e+00,  2.90108552e+00,\n",
       "        4.29376837e+00,  4.29637806e+00,  2.84031692e+00,  5.18031006e+00,\n",
       "        5.87099810e+00,  7.17960596e+00,  7.33880482e+00,  7.22512167e+00,\n",
       "        9.89222956e+00,  1.09292864e+01,  9.38897592e+00,  1.07911488e+01,\n",
       "        1.37253683e+01,  1.49116931e+01,  1.42233533e+01,  1.70311322e+01,\n",
       "        1.70796981e+01,  2.03929177e+01,  1.91172321e+01,  2.07938528e+01,\n",
       "        2.32658570e+01,  2.35565107e+01,  2.54719019e+01,  2.62168371e+01,\n",
       "        2.93416933e+01,  2.96698932e+01,  3.05945845e+01,  3.24996425e+01,\n",
       "        3.33291713e+01,  3.59631142e+01,  3.84346338e+01,  3.84772209e+01,\n",
       "        4.03349402e+01,  4.48484477e+01,  4.49899455e+01,  4.61714621e+01,\n",
       "        4.78006094e+01,  5.18658635e+01,  5.34403356e+01,  5.38403197e+01])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 2.5 画图看一下\n",
    "plt.plot(x_, y_data)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2.6 转为tensor \n",
    "x_train = torch.from_numpy(x_data).float()\n",
    "y_train = torch.from_numpy(y_data).float().unsqueeze(1)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 3 创建模型：2输入1输出 \n",
    "class PolyLinearRegression(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(PolyLinearRegression, self).__init__()\n",
    "        self.ploy = nn.Linear(2, 1)  \n",
    "\n",
    "    def forward(self, x):\n",
    "        out = self.ploy(x)\n",
    "        return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_and_test():\n",
    "    if torch.cuda.is_available():\n",
    "        device = torch.device(\"cuda:0\")\n",
    "        model = PolyLinearRegression()\n",
    "        model.to(device)\n",
    "    else:\n",
    "        device = torch.device(\"cpu\")\n",
    "        model = PolyLinearRegression()\n",
    "        model.to(device)\n",
    "\n",
    "    criterion = nn.MSELoss()\n",
    "    optimizer = torch.optim.SGD(model.parameters(), lr=0.001)\n",
    "\n",
    "    loss_list = []\n",
    "    for i in range(10000):\n",
    "        i += 1\n",
    "\n",
    "        inputs = x_train.to(device)\n",
    "        labels = y_train.to(device)\n",
    "\n",
    "        # 1 前向传播\n",
    "        output = model(inputs)\n",
    "        # 计算损失\n",
    "        loss = criterion(output, labels)\n",
    "\n",
    "        # 2 反向传播\n",
    "        # 梯度要清0\n",
    "        optimizer.zero_grad()\n",
    "        loss.backward()\n",
    "        # 更新权重参数\n",
    "        optimizer.step()\n",
    "\n",
    "        # 打印\n",
    "        if i % 10 == 0:\n",
    "            print(\"迭代次数：%d,  loss值：%.8f\" % (i, loss.item()))\n",
    "            loss_list.append(loss.item())\n",
    "\n",
    "    # 测试模式\n",
    "    model.eval()\n",
    "    predict = model(x_train).data.numpy()\n",
    "    plt.plot(x_, y_train)\n",
    "    plt.plot(x_, predict)\n",
    "    plt.show()\n",
    "\n",
    "    plt.plot(loss_list)\n",
    "    plt.show()\n",
    "    \n",
    "    # 查看参数\n",
    "    for p in model.parameters():\n",
    "        print(p)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "迭代次数：10,  loss值：18.50533676\n",
      "迭代次数：20,  loss值：10.88192368\n",
      "迭代次数：30,  loss值：8.04247379\n",
      "迭代次数：40,  loss值：6.02065754\n",
      "迭代次数：50,  loss值：4.57536602\n",
      "迭代次数：60,  loss值：3.54217911\n",
      "迭代次数：70,  loss值：2.80359244\n",
      "迭代次数：80,  loss值：2.27559948\n",
      "迭代次数：90,  loss值：1.89815068\n",
      "迭代次数：100,  loss值：1.62832379\n",
      "迭代次数：110,  loss值：1.43542659\n",
      "迭代次数：120,  loss值：1.29752588\n",
      "迭代次数：130,  loss值：1.19894004\n",
      "迭代次数：140,  loss值：1.12845707\n",
      "迭代次数：150,  loss值：1.07806528\n",
      "迭代次数：160,  loss值：1.04203510\n",
      "迭代次数：170,  loss值：1.01627159\n",
      "迭代次数：180,  loss值：0.99784768\n",
      "迭代次数：190,  loss值：0.98467088\n",
      "迭代次数：200,  loss值：0.97524458\n",
      "迭代次数：210,  loss值：0.96849942\n",
      "迭代次数：220,  loss值：0.96367162\n",
      "迭代次数：230,  loss值：0.96021414\n",
      "迭代次数：240,  loss值：0.95773673\n",
      "迭代次数：250,  loss值：0.95595950\n",
      "迭代次数：260,  loss值：0.95468307\n",
      "迭代次数：270,  loss值：0.95376533\n",
      "迭代次数：280,  loss值：0.95310342\n",
      "迭代次数：290,  loss值：0.95262432\n",
      "迭代次数：300,  loss值：0.95227653\n",
      "迭代次数：310,  loss值：0.95202255\n",
      "迭代次数：320,  loss值：0.95183593\n",
      "迭代次数：330,  loss值：0.95169711\n",
      "迭代次数：340,  loss值：0.95159274\n",
      "迭代次数：350,  loss值：0.95151353\n",
      "迭代次数：360,  loss值：0.95145166\n",
      "迭代次数：370,  loss值：0.95140254\n",
      "迭代次数：380,  loss值：0.95136279\n",
      "迭代次数：390,  loss值：0.95132977\n",
      "迭代次数：400,  loss值：0.95130134\n",
      "迭代次数：410,  loss值：0.95127678\n",
      "迭代次数：420,  loss值：0.95125479\n",
      "迭代次数：430,  loss值：0.95123440\n",
      "迭代次数：440,  loss值：0.95121551\n",
      "迭代次数：450,  loss值：0.95119816\n",
      "迭代次数：460,  loss值：0.95118153\n",
      "迭代次数：470,  loss值：0.95116585\n",
      "迭代次数：480,  loss值：0.95115036\n",
      "迭代次数：490,  loss值：0.95113546\n",
      "迭代次数：500,  loss值：0.95112067\n",
      "迭代次数：510,  loss值：0.95110661\n",
      "迭代次数：520,  loss值：0.95109260\n",
      "迭代次数：530,  loss值：0.95107925\n",
      "迭代次数：540,  loss值：0.95106620\n",
      "迭代次数：550,  loss值：0.95105338\n",
      "迭代次数：560,  loss值：0.95104086\n",
      "迭代次数：570,  loss值：0.95102805\n",
      "迭代次数：580,  loss值：0.95101577\n",
      "迭代次数：590,  loss值：0.95100379\n",
      "迭代次数：600,  loss值：0.95099163\n",
      "迭代次数：610,  loss值：0.95098048\n",
      "迭代次数：620,  loss值：0.95096892\n",
      "迭代次数：630,  loss值：0.95095772\n",
      "迭代次数：640,  loss值：0.95094663\n",
      "迭代次数：650,  loss值：0.95093602\n",
      "迭代次数：660,  loss值：0.95092553\n",
      "迭代次数：670,  loss值：0.95091480\n",
      "迭代次数：680,  loss值：0.95090461\n",
      "迭代次数：690,  loss值：0.95089459\n",
      "迭代次数：700,  loss值：0.95088452\n",
      "迭代次数：710,  loss值：0.95087492\n",
      "迭代次数：720,  loss值：0.95086527\n",
      "迭代次数：730,  loss值：0.95085585\n",
      "迭代次数：740,  loss值：0.95084673\n",
      "迭代次数：750,  loss值：0.95083743\n",
      "迭代次数：760,  loss值：0.95082879\n",
      "迭代次数：770,  loss值：0.95082009\n",
      "迭代次数：780,  loss值：0.95081145\n",
      "迭代次数：790,  loss值：0.95080298\n",
      "迭代次数：800,  loss值：0.95079494\n",
      "迭代次数：810,  loss值：0.95078683\n",
      "迭代次数：820,  loss值：0.95077860\n",
      "迭代次数：830,  loss值：0.95077068\n",
      "迭代次数：840,  loss值：0.95076323\n",
      "迭代次数：850,  loss值：0.95075530\n",
      "迭代次数：860,  loss值：0.95074815\n",
      "迭代次数：870,  loss值：0.95074081\n",
      "迭代次数：880,  loss值：0.95073390\n",
      "迭代次数：890,  loss值：0.95072645\n",
      "迭代次数：900,  loss值：0.95071965\n",
      "迭代次数：910,  loss值：0.95071256\n",
      "迭代次数：920,  loss值：0.95070618\n",
      "迭代次数：930,  loss值：0.95069927\n",
      "迭代次数：940,  loss值：0.95069307\n",
      "迭代次数：950,  loss值：0.95068669\n",
      "迭代次数：960,  loss值：0.95068032\n",
      "迭代次数：970,  loss值：0.95067418\n",
      "迭代次数：980,  loss值：0.95066833\n",
      "迭代次数：990,  loss值：0.95066237\n",
      "迭代次数：1000,  loss值：0.95065635\n",
      "迭代次数：1010,  loss值：0.95065069\n",
      "迭代次数：1020,  loss值：0.95064551\n",
      "迭代次数：1030,  loss值：0.95063955\n",
      "迭代次数：1040,  loss值：0.95063424\n",
      "迭代次数：1050,  loss值：0.95062894\n",
      "迭代次数：1060,  loss值：0.95062363\n",
      "迭代次数：1070,  loss值：0.95061827\n",
      "迭代次数：1080,  loss值：0.95061386\n",
      "迭代次数：1090,  loss值：0.95060867\n",
      "迭代次数：1100,  loss值：0.95060360\n",
      "迭代次数：1110,  loss值：0.95059884\n",
      "迭代次数：1120,  loss值：0.95059401\n",
      "迭代次数：1130,  loss值：0.95058960\n",
      "迭代次数：1140,  loss值：0.95058507\n",
      "迭代次数：1150,  loss值：0.95058060\n",
      "迭代次数：1160,  loss值：0.95057642\n",
      "迭代次数：1170,  loss值：0.95057213\n",
      "迭代次数：1180,  loss值：0.95056754\n",
      "迭代次数：1190,  loss值：0.95056331\n",
      "迭代次数：1200,  loss值：0.95055920\n",
      "迭代次数：1210,  loss值：0.95055550\n",
      "迭代次数：1220,  loss值：0.95055121\n",
      "迭代次数：1230,  loss值：0.95054746\n",
      "迭代次数：1240,  loss值：0.95054376\n",
      "迭代次数：1250,  loss值：0.95053983\n",
      "迭代次数：1260,  loss值：0.95053649\n",
      "迭代次数：1270,  loss值：0.95053303\n",
      "迭代次数：1280,  loss值：0.95052904\n",
      "迭代次数：1290,  loss值：0.95052564\n",
      "迭代次数：1300,  loss值：0.95052236\n",
      "迭代次数：1310,  loss值：0.95051897\n",
      "迭代次数：1320,  loss值：0.95051563\n",
      "迭代次数：1330,  loss值：0.95051253\n",
      "迭代次数：1340,  loss值：0.95050931\n",
      "迭代次数：1350,  loss值：0.95050627\n",
      "迭代次数：1360,  loss值：0.95050293\n",
      "迭代次数：1370,  loss值：0.95050001\n",
      "迭代次数：1380,  loss值：0.95049679\n",
      "迭代次数：1390,  loss值：0.95049393\n",
      "迭代次数：1400,  loss值：0.95049107\n",
      "迭代次数：1410,  loss值：0.95048845\n",
      "迭代次数：1420,  loss值：0.95048529\n",
      "迭代次数：1430,  loss值：0.95048302\n",
      "迭代次数：1440,  loss值：0.95048040\n",
      "迭代次数：1450,  loss值：0.95047784\n",
      "迭代次数：1460,  loss值：0.95047522\n",
      "迭代次数：1470,  loss值：0.95047253\n",
      "迭代次数：1480,  loss值：0.95047009\n",
      "迭代次数：1490,  loss值：0.95046759\n",
      "迭代次数：1500,  loss值：0.95046520\n",
      "迭代次数：1510,  loss值：0.95046300\n",
      "迭代次数：1520,  loss值：0.95046026\n",
      "迭代次数：1530,  loss值：0.95045847\n",
      "迭代次数：1540,  loss值：0.95045608\n",
      "迭代次数：1550,  loss值：0.95045382\n",
      "迭代次数：1560,  loss值：0.95045191\n",
      "迭代次数：1570,  loss值：0.95044971\n",
      "迭代次数：1580,  loss值：0.95044738\n",
      "迭代次数：1590,  loss值：0.95044541\n",
      "迭代次数：1600,  loss值：0.95044333\n",
      "迭代次数：1610,  loss值：0.95044118\n",
      "迭代次数：1620,  loss值：0.95043957\n",
      "迭代次数：1630,  loss值：0.95043737\n",
      "迭代次数：1640,  loss值：0.95043582\n",
      "迭代次数：1650,  loss值：0.95043373\n",
      "迭代次数：1660,  loss值：0.95043212\n",
      "迭代次数：1670,  loss值：0.95043033\n",
      "迭代次数：1680,  loss值：0.95042855\n",
      "迭代次数：1690,  loss值：0.95042711\n",
      "迭代次数：1700,  loss值：0.95042527\n",
      "迭代次数：1710,  loss值：0.95042336\n",
      "迭代次数：1720,  loss值：0.95042163\n",
      "迭代次数：1730,  loss值：0.95042014\n",
      "迭代次数：1740,  loss值：0.95041859\n",
      "迭代次数：1750,  loss值：0.95041716\n",
      "迭代次数：1760,  loss值：0.95041543\n",
      "迭代次数：1770,  loss值：0.95041430\n",
      "迭代次数：1780,  loss值：0.95041287\n",
      "迭代次数：1790,  loss值：0.95041156\n",
      "迭代次数：1800,  loss值：0.95040983\n",
      "迭代次数：1810,  loss值：0.95040816\n",
      "迭代次数：1820,  loss值：0.95040721\n",
      "迭代次数：1830,  loss值：0.95040578\n",
      "迭代次数：1840,  loss值：0.95040429\n",
      "迭代次数：1850,  loss值：0.95040339\n",
      "迭代次数：1860,  loss值：0.95040178\n",
      "迭代次数：1870,  loss值：0.95040095\n",
      "迭代次数：1880,  loss值：0.95039940\n",
      "迭代次数：1890,  loss值：0.95039827\n",
      "迭代次数：1900,  loss值：0.95039713\n",
      "迭代次数：1910,  loss值：0.95039582\n",
      "迭代次数：1920,  loss值：0.95039457\n",
      "迭代次数：1930,  loss值：0.95039397\n",
      "迭代次数：1940,  loss值：0.95039237\n",
      "迭代次数：1950,  loss值：0.95039147\n",
      "迭代次数：1960,  loss值：0.95039022\n",
      "迭代次数：1970,  loss值：0.95038974\n",
      "迭代次数：1980,  loss值：0.95038831\n",
      "迭代次数：1990,  loss值：0.95038724\n",
      "迭代次数：2000,  loss值：0.95038640\n",
      "迭代次数：2010,  loss值：0.95038539\n",
      "迭代次数：2020,  loss值：0.95038468\n",
      "迭代次数：2030,  loss值：0.95038319\n",
      "迭代次数：2040,  loss值：0.95038265\n",
      "迭代次数：2050,  loss值：0.95038188\n",
      "迭代次数：2060,  loss值：0.95038074\n",
      "迭代次数：2070,  loss值：0.95037991\n",
      "迭代次数：2080,  loss值：0.95037895\n",
      "迭代次数：2090,  loss值：0.95037788\n",
      "迭代次数：2100,  loss值：0.95037723\n",
      "迭代次数：2110,  loss值：0.95037639\n",
      "迭代次数：2120,  loss值：0.95037591\n",
      "迭代次数：2130,  loss值：0.95037490\n",
      "迭代次数：2140,  loss值：0.95037413\n",
      "迭代次数：2150,  loss值：0.95037323\n",
      "迭代次数：2160,  loss值：0.95037252\n",
      "迭代次数：2170,  loss值：0.95037186\n",
      "迭代次数：2180,  loss值：0.95037091\n",
      "迭代次数：2190,  loss值：0.95037049\n",
      "迭代次数：2200,  loss值：0.95037013\n",
      "迭代次数：2210,  loss值：0.95036906\n",
      "迭代次数：2220,  loss值：0.95036817\n",
      "迭代次数：2230,  loss值：0.95036781\n",
      "迭代次数：2240,  loss值：0.95036697\n",
      "迭代次数：2250,  loss值：0.95036644\n",
      "迭代次数：2260,  loss值：0.95036608\n",
      "迭代次数：2270,  loss值：0.95036536\n",
      "迭代次数：2280,  loss值：0.95036459\n",
      "迭代次数：2290,  loss值：0.95036399\n",
      "迭代次数：2300,  loss值：0.95036346\n",
      "迭代次数：2310,  loss值：0.95036262\n",
      "迭代次数：2320,  loss值：0.95036221\n",
      "迭代次数：2330,  loss值：0.95036167\n",
      "迭代次数：2340,  loss值：0.95036131\n",
      "迭代次数：2350,  loss值：0.95036066\n",
      "迭代次数：2360,  loss值：0.95036030\n",
      "迭代次数：2370,  loss值：0.95035970\n",
      "迭代次数：2380,  loss值：0.95035934\n",
      "迭代次数：2390,  loss值：0.95035875\n",
      "迭代次数：2400,  loss值：0.95035815\n",
      "迭代次数：2410,  loss值：0.95035791\n",
      "迭代次数：2420,  loss值：0.95035750\n",
      "迭代次数：2430,  loss值：0.95035678\n",
      "迭代次数：2440,  loss值：0.95035630\n",
      "迭代次数：2450,  loss值：0.95035607\n",
      "迭代次数：2460,  loss值：0.95035523\n",
      "迭代次数：2470,  loss值：0.95035511\n",
      "迭代次数：2480,  loss值：0.95035475\n",
      "迭代次数：2490,  loss值：0.95035416\n",
      "迭代次数：2500,  loss值：0.95035392\n",
      "迭代次数：2510,  loss值：0.95035332\n",
      "迭代次数：2520,  loss值：0.95035297\n",
      "迭代次数：2530,  loss值：0.95035261\n",
      "迭代次数：2540,  loss值：0.95035249\n",
      "迭代次数：2550,  loss值：0.95035172\n",
      "迭代次数：2560,  loss值：0.95035136\n",
      "迭代次数：2570,  loss值：0.95035136\n",
      "迭代次数：2580,  loss值：0.95035052\n",
      "迭代次数：2590,  loss值：0.95035028\n",
      "迭代次数：2600,  loss值：0.95034993\n",
      "迭代次数：2610,  loss值：0.95034981\n",
      "迭代次数：2620,  loss值：0.95034921\n",
      "迭代次数：2630,  loss值：0.95034939\n",
      "迭代次数：2640,  loss值：0.95034868\n",
      "迭代次数：2650,  loss值：0.95034850\n",
      "迭代次数：2660,  loss值：0.95034772\n",
      "迭代次数：2670,  loss值：0.95034784\n",
      "迭代次数：2680,  loss值：0.95034748\n",
      "迭代次数：2690,  loss值：0.95034736\n",
      "迭代次数：2700,  loss值：0.95034683\n",
      "迭代次数：2710,  loss值：0.95034677\n",
      "迭代次数：2720,  loss值：0.95034659\n",
      "迭代次数：2730,  loss值：0.95034611\n",
      "迭代次数：2740,  loss值：0.95034570\n",
      "迭代次数：2750,  loss值：0.95034570\n",
      "迭代次数：2760,  loss值：0.95034546\n",
      "迭代次数：2770,  loss值：0.95034516\n",
      "迭代次数：2780,  loss值：0.95034468\n",
      "迭代次数：2790,  loss值：0.95034474\n",
      "迭代次数：2800,  loss值：0.95034444\n",
      "迭代次数：2810,  loss值：0.95034403\n",
      "迭代次数：2820,  loss值：0.95034397\n",
      "迭代次数：2830,  loss值：0.95034355\n",
      "迭代次数：2840,  loss值：0.95034367\n",
      "迭代次数：2850,  loss值：0.95034319\n",
      "迭代次数：2860,  loss值：0.95034301\n",
      "迭代次数：2870,  loss值：0.95034254\n",
      "迭代次数：2880,  loss值：0.95034248\n",
      "迭代次数：2890,  loss值：0.95034236\n",
      "迭代次数：2900,  loss值：0.95034230\n",
      "迭代次数：2910,  loss值：0.95034182\n",
      "迭代次数：2920,  loss值：0.95034158\n",
      "迭代次数：2930,  loss值：0.95034158\n",
      "迭代次数：2940,  loss值：0.95034152\n",
      "迭代次数：2950,  loss值：0.95034134\n",
      "迭代次数：2960,  loss值：0.95034099\n",
      "迭代次数：2970,  loss值：0.95034087\n",
      "迭代次数：2980,  loss值：0.95034069\n",
      "迭代次数：2990,  loss值：0.95034069\n",
      "迭代次数：3000,  loss值：0.95034051\n",
      "迭代次数：3010,  loss值：0.95034009\n",
      "迭代次数：3020,  loss值：0.95033991\n",
      "迭代次数：3030,  loss值：0.95033962\n",
      "迭代次数：3040,  loss值：0.95033962\n",
      "迭代次数：3050,  loss值：0.95033956\n",
      "迭代次数：3060,  loss值：0.95033967\n",
      "迭代次数：3070,  loss值：0.95033914\n",
      "迭代次数：3080,  loss值：0.95033920\n",
      "迭代次数：3090,  loss值：0.95033914\n",
      "迭代次数：3100,  loss值：0.95033860\n",
      "迭代次数：3110,  loss值：0.95033860\n",
      "迭代次数：3120,  loss值：0.95033860\n",
      "迭代次数：3130,  loss值：0.95033836\n",
      "迭代次数：3140,  loss值：0.95033824\n",
      "迭代次数：3150,  loss值：0.95033830\n",
      "迭代次数：3160,  loss值：0.95033783\n",
      "迭代次数：3170,  loss值：0.95033795\n",
      "迭代次数：3180,  loss值：0.95033795\n",
      "迭代次数：3190,  loss值：0.95033789\n",
      "迭代次数：3200,  loss值：0.95033735\n",
      "迭代次数：3210,  loss值：0.95033711\n",
      "迭代次数：3220,  loss值：0.95033699\n",
      "迭代次数：3230,  loss值：0.95033693\n",
      "迭代次数：3240,  loss值：0.95033723\n",
      "迭代次数：3250,  loss值：0.95033693\n",
      "迭代次数：3260,  loss值：0.95033687\n",
      "迭代次数：3270,  loss值：0.95033658\n",
      "迭代次数：3280,  loss值：0.95033675\n",
      "迭代次数：3290,  loss值：0.95033658\n",
      "迭代次数：3300,  loss值：0.95033646\n",
      "迭代次数：3310,  loss值：0.95033640\n",
      "迭代次数：3320,  loss值：0.95033628\n",
      "迭代次数：3330,  loss值：0.95033616\n",
      "迭代次数：3340,  loss值：0.95033592\n",
      "迭代次数：3350,  loss值：0.95033574\n",
      "迭代次数：3360,  loss值：0.95033598\n",
      "迭代次数：3370,  loss值：0.95033592\n",
      "迭代次数：3380,  loss值：0.95033592\n",
      "迭代次数：3390,  loss值：0.95033568\n",
      "迭代次数：3400,  loss值：0.95033550\n",
      "迭代次数：3410,  loss值：0.95033562\n",
      "迭代次数：3420,  loss值：0.95033550\n",
      "迭代次数：3430,  loss值：0.95033532\n",
      "迭代次数：3440,  loss值：0.95033514\n",
      "迭代次数：3450,  loss值：0.95033520\n",
      "迭代次数：3460,  loss值：0.95033526\n",
      "迭代次数：3470,  loss值：0.95033509\n",
      "迭代次数：3480,  loss值：0.95033497\n",
      "迭代次数：3490,  loss值：0.95033503\n",
      "迭代次数：3500,  loss值：0.95033491\n",
      "迭代次数：3510,  loss值：0.95033473\n",
      "迭代次数：3520,  loss值：0.95033467\n",
      "迭代次数：3530,  loss值：0.95033467\n",
      "迭代次数：3540,  loss值：0.95033461\n",
      "迭代次数：3550,  loss值：0.95033449\n",
      "迭代次数：3560,  loss值：0.95033461\n",
      "迭代次数：3570,  loss值：0.95033425\n",
      "迭代次数：3580,  loss值：0.95033419\n",
      "迭代次数：3590,  loss值：0.95033401\n",
      "迭代次数：3600,  loss值：0.95033401\n",
      "迭代次数：3610,  loss值：0.95033437\n",
      "迭代次数：3620,  loss值：0.95033401\n",
      "迭代次数：3630,  loss值：0.95033419\n",
      "迭代次数：3640,  loss值：0.95033389\n",
      "迭代次数：3650,  loss值：0.95033395\n",
      "迭代次数：3660,  loss值：0.95033389\n",
      "迭代次数：3670,  loss值：0.95033365\n",
      "迭代次数：3680,  loss值：0.95033383\n",
      "迭代次数：3690,  loss值：0.95033377\n",
      "迭代次数：3700,  loss值：0.95033354\n",
      "迭代次数：3710,  loss值：0.95033360\n",
      "迭代次数：3720,  loss值：0.95033318\n",
      "迭代次数：3730,  loss值：0.95033365\n",
      "迭代次数：3740,  loss值：0.95033348\n",
      "迭代次数：3750,  loss值：0.95033360\n",
      "迭代次数：3760,  loss值：0.95033330\n",
      "迭代次数：3770,  loss值：0.95033330\n",
      "迭代次数：3780,  loss值：0.95033318\n",
      "迭代次数：3790,  loss值：0.95033348\n",
      "迭代次数：3800,  loss值：0.95033324\n",
      "迭代次数：3810,  loss值：0.95033342\n",
      "迭代次数：3820,  loss值：0.95033318\n",
      "迭代次数：3830,  loss值：0.95033300\n",
      "迭代次数：3840,  loss值：0.95033330\n",
      "迭代次数：3850,  loss值：0.95033300\n",
      "迭代次数：3860,  loss值：0.95033300\n",
      "迭代次数：3870,  loss值：0.95033300\n",
      "迭代次数：3880,  loss值：0.95033288\n",
      "迭代次数：3890,  loss值：0.95033294\n",
      "迭代次数：3900,  loss值：0.95033294\n",
      "迭代次数：3910,  loss值：0.95033282\n",
      "迭代次数：3920,  loss值：0.95033270\n",
      "迭代次数：3930,  loss值：0.95033294\n",
      "迭代次数：3940,  loss值：0.95033294\n",
      "迭代次数：3950,  loss值：0.95033276\n",
      "迭代次数：3960,  loss值：0.95033252\n",
      "迭代次数：3970,  loss值：0.95033240\n",
      "迭代次数：3980,  loss值：0.95033264\n",
      "迭代次数：3990,  loss值：0.95033246\n",
      "迭代次数：4000,  loss值：0.95033240\n",
      "迭代次数：4010,  loss值：0.95033252\n",
      "迭代次数：4020,  loss值：0.95033234\n",
      "迭代次数：4030,  loss值：0.95033258\n",
      "迭代次数：4040,  loss值：0.95033264\n",
      "迭代次数：4050,  loss值：0.95033246\n",
      "迭代次数：4060,  loss值：0.95033240\n",
      "迭代次数：4070,  loss值：0.95033222\n",
      "迭代次数：4080,  loss值：0.95033222\n",
      "迭代次数：4090,  loss值：0.95033234\n",
      "迭代次数：4100,  loss值：0.95033228\n",
      "迭代次数：4110,  loss值：0.95033234\n",
      "迭代次数：4120,  loss值：0.95033216\n",
      "迭代次数：4130,  loss值：0.95033187\n",
      "迭代次数：4140,  loss值：0.95033216\n",
      "迭代次数：4150,  loss值：0.95033216\n",
      "迭代次数：4160,  loss值：0.95033199\n",
      "迭代次数：4170,  loss值：0.95033228\n",
      "迭代次数：4180,  loss值：0.95033187\n",
      "迭代次数：4190,  loss值：0.95033205\n",
      "迭代次数：4200,  loss值：0.95033193\n",
      "迭代次数：4210,  loss值：0.95033205\n",
      "迭代次数：4220,  loss值：0.95033211\n",
      "迭代次数：4230,  loss值：0.95033211\n",
      "迭代次数：4240,  loss值：0.95033193\n",
      "迭代次数：4250,  loss值：0.95033175\n",
      "迭代次数：4260,  loss值：0.95033163\n",
      "迭代次数：4270,  loss值：0.95033175\n",
      "迭代次数：4280,  loss值：0.95033187\n",
      "迭代次数：4290,  loss值：0.95033187\n",
      "迭代次数：4300,  loss值：0.95033187\n",
      "迭代次数：4310,  loss值：0.95033193\n",
      "迭代次数：4320,  loss值：0.95033187\n",
      "迭代次数：4330,  loss值：0.95033169\n",
      "迭代次数：4340,  loss值：0.95033175\n",
      "迭代次数：4350,  loss值：0.95033187\n",
      "迭代次数：4360,  loss值：0.95033181\n",
      "迭代次数：4370,  loss值：0.95033181\n",
      "迭代次数：4380,  loss值：0.95033151\n",
      "迭代次数：4390,  loss值：0.95033193\n",
      "迭代次数：4400,  loss值：0.95033169\n",
      "迭代次数：4410,  loss值：0.95033193\n",
      "迭代次数：4420,  loss值：0.95033193\n",
      "迭代次数：4430,  loss值：0.95033163\n",
      "迭代次数：4440,  loss值：0.95033163\n",
      "迭代次数：4450,  loss值：0.95033175\n",
      "迭代次数：4460,  loss值：0.95033175\n",
      "迭代次数：4470,  loss值：0.95033175\n",
      "迭代次数：4480,  loss值：0.95033175\n",
      "迭代次数：4490,  loss值：0.95033169\n",
      "迭代次数：4500,  loss值：0.95033169\n",
      "迭代次数：4510,  loss值：0.95033175\n",
      "迭代次数：4520,  loss值：0.95033151\n",
      "迭代次数：4530,  loss值：0.95033133\n",
      "迭代次数：4540,  loss值：0.95033157\n",
      "迭代次数：4550,  loss值：0.95033139\n",
      "迭代次数：4560,  loss值：0.95033151\n",
      "迭代次数：4570,  loss值：0.95033121\n",
      "迭代次数：4580,  loss值：0.95033157\n",
      "迭代次数：4590,  loss值：0.95033157\n",
      "迭代次数：4600,  loss值：0.95033115\n",
      "迭代次数：4610,  loss值：0.95033151\n",
      "迭代次数：4620,  loss值：0.95033145\n",
      "迭代次数：4630,  loss值：0.95033187\n",
      "迭代次数：4640,  loss值：0.95033145\n",
      "迭代次数：4650,  loss值：0.95033145\n",
      "迭代次数：4660,  loss值：0.95033169\n",
      "迭代次数：4670,  loss值：0.95033145\n",
      "迭代次数：4680,  loss值：0.95033145\n",
      "迭代次数：4690,  loss值：0.95033133\n",
      "迭代次数：4700,  loss值：0.95033169\n",
      "迭代次数：4710,  loss值：0.95033151\n",
      "迭代次数：4720,  loss值：0.95033121\n",
      "迭代次数：4730,  loss值：0.95033121\n",
      "迭代次数：4740,  loss值：0.95033133\n",
      "迭代次数：4750,  loss值：0.95033145\n",
      "迭代次数：4760,  loss值：0.95033121\n",
      "迭代次数：4770,  loss值：0.95033109\n",
      "迭代次数：4780,  loss值：0.95033121\n",
      "迭代次数：4790,  loss值：0.95033139\n",
      "迭代次数：4800,  loss值：0.95033121\n",
      "迭代次数：4810,  loss值：0.95033097\n",
      "迭代次数：4820,  loss值：0.95033133\n",
      "迭代次数：4830,  loss值：0.95033121\n",
      "迭代次数：4840,  loss值：0.95033121\n",
      "迭代次数：4850,  loss值：0.95033157\n",
      "迭代次数：4860,  loss值：0.95033133\n",
      "迭代次数：4870,  loss值：0.95033133\n",
      "迭代次数：4880,  loss值：0.95033127\n",
      "迭代次数：4890,  loss值：0.95033151\n",
      "迭代次数：4900,  loss值：0.95033133\n",
      "迭代次数：4910,  loss值：0.95033139\n",
      "迭代次数：4920,  loss值：0.95033103\n",
      "迭代次数：4930,  loss值：0.95033115\n",
      "迭代次数：4940,  loss值：0.95033109\n",
      "迭代次数：4950,  loss值：0.95033103\n",
      "迭代次数：4960,  loss值：0.95033139\n",
      "迭代次数：4970,  loss值：0.95033139\n",
      "迭代次数：4980,  loss值：0.95033139\n",
      "迭代次数：4990,  loss值：0.95033103\n",
      "迭代次数：5000,  loss值：0.95033115\n",
      "迭代次数：5010,  loss值：0.95033097\n",
      "迭代次数：5020,  loss值：0.95033073\n",
      "迭代次数：5030,  loss值：0.95033091\n",
      "迭代次数：5040,  loss值：0.95033127\n",
      "迭代次数：5050,  loss值：0.95033097\n",
      "迭代次数：5060,  loss值：0.95033103\n",
      "迭代次数：5070,  loss值：0.95033097\n",
      "迭代次数：5080,  loss值：0.95033121\n",
      "迭代次数：5090,  loss值：0.95033091\n",
      "迭代次数：5100,  loss值：0.95033115\n",
      "迭代次数：5110,  loss值：0.95033121\n",
      "迭代次数：5120,  loss值：0.95033091\n",
      "迭代次数：5130,  loss值：0.95033103\n",
      "迭代次数：5140,  loss值：0.95033139\n",
      "迭代次数：5150,  loss值：0.95033103\n",
      "迭代次数：5160,  loss值：0.95033097\n",
      "迭代次数：5170,  loss值：0.95033097\n",
      "迭代次数：5180,  loss值：0.95033115\n",
      "迭代次数：5190,  loss值：0.95033091\n",
      "迭代次数：5200,  loss值：0.95033103\n",
      "迭代次数：5210,  loss值：0.95033115\n",
      "迭代次数：5220,  loss值：0.95033127\n",
      "迭代次数：5230,  loss值：0.95033109\n",
      "迭代次数：5240,  loss值：0.95033133\n",
      "迭代次数：5250,  loss值：0.95033109\n",
      "迭代次数：5260,  loss值：0.95033109\n",
      "迭代次数：5270,  loss值：0.95033139\n",
      "迭代次数：5280,  loss值：0.95033145\n",
      "迭代次数：5290,  loss值：0.95033103\n",
      "迭代次数：5300,  loss值：0.95033133\n",
      "迭代次数：5310,  loss值：0.95033127\n",
      "迭代次数：5320,  loss值：0.95033097\n",
      "迭代次数：5330,  loss值：0.95033121\n",
      "迭代次数：5340,  loss值：0.95033103\n",
      "迭代次数：5350,  loss值：0.95033103\n",
      "迭代次数：5360,  loss值：0.95033085\n",
      "迭代次数：5370,  loss值：0.95033115\n",
      "迭代次数：5380,  loss值：0.95033073\n",
      "迭代次数：5390,  loss值：0.95033133\n",
      "迭代次数：5400,  loss值：0.95033115\n",
      "迭代次数：5410,  loss值：0.95033115\n",
      "迭代次数：5420,  loss值：0.95033091\n",
      "迭代次数：5430,  loss值：0.95033085\n",
      "迭代次数：5440,  loss值：0.95033085\n",
      "迭代次数：5450,  loss值：0.95033073\n",
      "迭代次数：5460,  loss值：0.95033091\n",
      "迭代次数：5470,  loss值：0.95033091\n",
      "迭代次数：5480,  loss值：0.95033097\n",
      "迭代次数：5490,  loss值：0.95033091\n",
      "迭代次数：5500,  loss值：0.95033085\n",
      "迭代次数：5510,  loss值：0.95033085\n",
      "迭代次数：5520,  loss值：0.95033079\n",
      "迭代次数：5530,  loss值：0.95033073\n",
      "迭代次数：5540,  loss值：0.95033097\n",
      "迭代次数：5550,  loss值：0.95033103\n",
      "迭代次数：5560,  loss值：0.95033085\n",
      "迭代次数：5570,  loss值：0.95033091\n",
      "迭代次数：5580,  loss值：0.95033091\n",
      "迭代次数：5590,  loss值：0.95033097\n",
      "迭代次数：5600,  loss值：0.95033079\n",
      "迭代次数：5610,  loss值：0.95033079\n",
      "迭代次数：5620,  loss值：0.95033091\n",
      "迭代次数：5630,  loss值：0.95033097\n",
      "迭代次数：5640,  loss值：0.95033085\n",
      "迭代次数：5650,  loss值：0.95033097\n",
      "迭代次数：5660,  loss值：0.95033109\n",
      "迭代次数：5670,  loss值：0.95033109\n",
      "迭代次数：5680,  loss值：0.95033103\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "迭代次数：5690,  loss值：0.95033109\n",
      "迭代次数：5700,  loss值：0.95033115\n",
      "迭代次数：5710,  loss值：0.95033085\n",
      "迭代次数：5720,  loss值：0.95033073\n",
      "迭代次数：5730,  loss值：0.95033103\n",
      "迭代次数：5740,  loss值：0.95033079\n",
      "迭代次数：5750,  loss值：0.95033115\n",
      "迭代次数：5760,  loss值：0.95033121\n",
      "迭代次数：5770,  loss值：0.95033115\n",
      "迭代次数：5780,  loss值：0.95033115\n",
      "迭代次数：5790,  loss值：0.95033103\n",
      "迭代次数：5800,  loss值：0.95033097\n",
      "迭代次数：5810,  loss值：0.95033127\n",
      "迭代次数：5820,  loss值：0.95033097\n",
      "迭代次数：5830,  loss值：0.95033097\n",
      "迭代次数：5840,  loss值：0.95033079\n",
      "迭代次数：5850,  loss值：0.95033127\n",
      "迭代次数：5860,  loss值：0.95033109\n",
      "迭代次数：5870,  loss值：0.95033091\n",
      "迭代次数：5880,  loss值：0.95033127\n",
      "迭代次数：5890,  loss值：0.95033091\n",
      "迭代次数：5900,  loss值：0.95033133\n",
      "迭代次数：5910,  loss值：0.95033091\n",
      "迭代次数：5920,  loss值：0.95033091\n",
      "迭代次数：5930,  loss值：0.95033121\n",
      "迭代次数：5940,  loss值：0.95033103\n",
      "迭代次数：5950,  loss值：0.95033073\n",
      "迭代次数：5960,  loss值：0.95033103\n",
      "迭代次数：5970,  loss值：0.95033091\n",
      "迭代次数：5980,  loss值：0.95033079\n",
      "迭代次数：5990,  loss值：0.95033073\n",
      "迭代次数：6000,  loss值：0.95033091\n",
      "迭代次数：6010,  loss值：0.95033085\n",
      "迭代次数：6020,  loss值：0.95033067\n",
      "迭代次数：6030,  loss值：0.95033085\n",
      "迭代次数：6040,  loss值：0.95033079\n",
      "迭代次数：6050,  loss值：0.95033062\n",
      "迭代次数：6060,  loss值：0.95033109\n",
      "迭代次数：6070,  loss值：0.95033073\n",
      "迭代次数：6080,  loss值：0.95033085\n",
      "迭代次数：6090,  loss值：0.95033079\n",
      "迭代次数：6100,  loss值：0.95033056\n",
      "迭代次数：6110,  loss值：0.95033067\n",
      "迭代次数：6120,  loss值：0.95033073\n",
      "迭代次数：6130,  loss值：0.95033091\n",
      "迭代次数：6140,  loss值：0.95033091\n",
      "迭代次数：6150,  loss值：0.95033079\n",
      "迭代次数：6160,  loss值：0.95033067\n",
      "迭代次数：6170,  loss值：0.95033091\n",
      "迭代次数：6180,  loss值：0.95033097\n",
      "迭代次数：6190,  loss值：0.95033097\n",
      "迭代次数：6200,  loss值：0.95033073\n",
      "迭代次数：6210,  loss值：0.95033097\n",
      "迭代次数：6220,  loss值：0.95033085\n",
      "迭代次数：6230,  loss值：0.95033091\n",
      "迭代次数：6240,  loss值：0.95033085\n",
      "迭代次数：6250,  loss值：0.95033085\n",
      "迭代次数：6260,  loss值：0.95033079\n",
      "迭代次数：6270,  loss值：0.95033115\n",
      "迭代次数：6280,  loss值：0.95033073\n",
      "迭代次数：6290,  loss值：0.95033073\n",
      "迭代次数：6300,  loss值：0.95033085\n",
      "迭代次数：6310,  loss值：0.95033085\n",
      "迭代次数：6320,  loss值：0.95033121\n",
      "迭代次数：6330,  loss值：0.95033127\n",
      "迭代次数：6340,  loss值：0.95033079\n",
      "迭代次数：6350,  loss值：0.95033091\n",
      "迭代次数：6360,  loss值：0.95033097\n",
      "迭代次数：6370,  loss值：0.95033097\n",
      "迭代次数：6380,  loss值：0.95033085\n",
      "迭代次数：6390,  loss值：0.95033073\n",
      "迭代次数：6400,  loss值：0.95033103\n",
      "迭代次数：6410,  loss值：0.95033079\n",
      "迭代次数：6420,  loss值：0.95033085\n",
      "迭代次数：6430,  loss值：0.95033103\n",
      "迭代次数：6440,  loss值：0.95033091\n",
      "迭代次数：6450,  loss值：0.95033103\n",
      "迭代次数：6460,  loss值：0.95033073\n",
      "迭代次数：6470,  loss值：0.95033085\n",
      "迭代次数：6480,  loss值：0.95033085\n",
      "迭代次数：6490,  loss值：0.95033103\n",
      "迭代次数：6500,  loss值：0.95033097\n",
      "迭代次数：6510,  loss值：0.95033079\n",
      "迭代次数：6520,  loss值：0.95033115\n",
      "迭代次数：6530,  loss值：0.95033115\n",
      "迭代次数：6540,  loss值：0.95033109\n",
      "迭代次数：6550,  loss值：0.95033121\n",
      "迭代次数：6560,  loss值：0.95033115\n",
      "迭代次数：6570,  loss值：0.95033127\n",
      "迭代次数：6580,  loss值：0.95033127\n",
      "迭代次数：6590,  loss值：0.95033079\n",
      "迭代次数：6600,  loss值：0.95033103\n",
      "迭代次数：6610,  loss值：0.95033091\n",
      "迭代次数：6620,  loss值：0.95033103\n",
      "迭代次数：6630,  loss值：0.95033115\n",
      "迭代次数：6640,  loss值：0.95033115\n",
      "迭代次数：6650,  loss值：0.95033085\n",
      "迭代次数：6660,  loss值：0.95033085\n",
      "迭代次数：6670,  loss值：0.95033115\n",
      "迭代次数：6680,  loss值：0.95033115\n",
      "迭代次数：6690,  loss值：0.95033073\n",
      "迭代次数：6700,  loss值：0.95033121\n",
      "迭代次数：6710,  loss值：0.95033115\n",
      "迭代次数：6720,  loss值：0.95033103\n",
      "迭代次数：6730,  loss值：0.95033109\n",
      "迭代次数：6740,  loss值：0.95033121\n",
      "迭代次数：6750,  loss值：0.95033115\n",
      "迭代次数：6760,  loss值：0.95033079\n",
      "迭代次数：6770,  loss值：0.95033103\n",
      "迭代次数：6780,  loss值：0.95033109\n",
      "迭代次数：6790,  loss值：0.95033073\n",
      "迭代次数：6800,  loss值：0.95033109\n",
      "迭代次数：6810,  loss值：0.95033103\n",
      "迭代次数：6820,  loss值：0.95033103\n",
      "迭代次数：6830,  loss值：0.95033085\n",
      "迭代次数：6840,  loss值：0.95033079\n",
      "迭代次数：6850,  loss值：0.95033109\n",
      "迭代次数：6860,  loss值：0.95033103\n",
      "迭代次数：6870,  loss值：0.95033097\n",
      "迭代次数：6880,  loss值：0.95033085\n",
      "迭代次数：6890,  loss值：0.95033091\n",
      "迭代次数：6900,  loss值：0.95033103\n",
      "迭代次数：6910,  loss值：0.95033097\n",
      "迭代次数：6920,  loss值：0.95033097\n",
      "迭代次数：6930,  loss值：0.95033085\n",
      "迭代次数：6940,  loss值：0.95033109\n",
      "迭代次数：6950,  loss值：0.95033115\n",
      "迭代次数：6960,  loss值：0.95033085\n",
      "迭代次数：6970,  loss值：0.95033091\n",
      "迭代次数：6980,  loss值：0.95033097\n",
      "迭代次数：6990,  loss值：0.95033097\n",
      "迭代次数：7000,  loss值：0.95033127\n",
      "迭代次数：7010,  loss值：0.95033085\n",
      "迭代次数：7020,  loss值：0.95033079\n",
      "迭代次数：7030,  loss值：0.95033073\n",
      "迭代次数：7040,  loss值：0.95033097\n",
      "迭代次数：7050,  loss值：0.95033097\n",
      "迭代次数：7060,  loss值：0.95033091\n",
      "迭代次数：7070,  loss值：0.95033056\n",
      "迭代次数：7080,  loss值：0.95033056\n",
      "迭代次数：7090,  loss值：0.95033050\n",
      "迭代次数：7100,  loss值：0.95033079\n",
      "迭代次数：7110,  loss值：0.95033073\n",
      "迭代次数：7120,  loss值：0.95033079\n",
      "迭代次数：7130,  loss值：0.95033062\n",
      "迭代次数：7140,  loss值：0.95033067\n",
      "迭代次数：7150,  loss值：0.95033062\n",
      "迭代次数：7160,  loss值：0.95033067\n",
      "迭代次数：7170,  loss值：0.95033115\n",
      "迭代次数：7180,  loss值：0.95033067\n",
      "迭代次数：7190,  loss值：0.95033073\n",
      "迭代次数：7200,  loss值：0.95033056\n",
      "迭代次数：7210,  loss值：0.95033091\n",
      "迭代次数：7220,  loss值：0.95033085\n",
      "迭代次数：7230,  loss值：0.95033091\n",
      "迭代次数：7240,  loss值：0.95033062\n",
      "迭代次数：7250,  loss值：0.95033085\n",
      "迭代次数：7260,  loss值：0.95033073\n",
      "迭代次数：7270,  loss值：0.95033085\n",
      "迭代次数：7280,  loss值：0.95033079\n",
      "迭代次数：7290,  loss值：0.95033091\n",
      "迭代次数：7300,  loss值：0.95033067\n",
      "迭代次数：7310,  loss值：0.95033067\n",
      "迭代次数：7320,  loss值：0.95033067\n",
      "迭代次数：7330,  loss值：0.95033056\n",
      "迭代次数：7340,  loss值：0.95033115\n",
      "迭代次数：7350,  loss值：0.95033056\n",
      "迭代次数：7360,  loss值：0.95033056\n",
      "迭代次数：7370,  loss值：0.95033050\n",
      "迭代次数：7380,  loss值：0.95033085\n",
      "迭代次数：7390,  loss值：0.95033097\n",
      "迭代次数：7400,  loss值：0.95033103\n",
      "迭代次数：7410,  loss值：0.95033091\n",
      "迭代次数：7420,  loss值：0.95033097\n",
      "迭代次数：7430,  loss值：0.95033091\n",
      "迭代次数：7440,  loss值：0.95033085\n",
      "迭代次数：7450,  loss值：0.95033085\n",
      "迭代次数：7460,  loss值：0.95033079\n",
      "迭代次数：7470,  loss值：0.95033067\n",
      "迭代次数：7480,  loss值：0.95033073\n",
      "迭代次数：7490,  loss值：0.95033062\n",
      "迭代次数：7500,  loss值：0.95033067\n",
      "迭代次数：7510,  loss值：0.95033073\n",
      "迭代次数：7520,  loss值：0.95033079\n",
      "迭代次数：7530,  loss值：0.95033085\n",
      "迭代次数：7540,  loss值：0.95033079\n",
      "迭代次数：7550,  loss值：0.95033085\n",
      "迭代次数：7560,  loss值：0.95033085\n",
      "迭代次数：7570,  loss值：0.95033091\n",
      "迭代次数：7580,  loss值：0.95033073\n",
      "迭代次数：7590,  loss值：0.95033067\n",
      "迭代次数：7600,  loss值：0.95033067\n",
      "迭代次数：7610,  loss值：0.95033109\n",
      "迭代次数：7620,  loss值：0.95033097\n",
      "迭代次数：7630,  loss值：0.95033073\n",
      "迭代次数：7640,  loss值：0.95033062\n",
      "迭代次数：7650,  loss值：0.95033062\n",
      "迭代次数：7660,  loss值：0.95033079\n",
      "迭代次数：7670,  loss值：0.95033085\n",
      "迭代次数：7680,  loss值：0.95033073\n",
      "迭代次数：7690,  loss值：0.95033109\n",
      "迭代次数：7700,  loss值：0.95033109\n",
      "迭代次数：7710,  loss值：0.95033103\n",
      "迭代次数：7720,  loss值：0.95033079\n",
      "迭代次数：7730,  loss值：0.95033079\n",
      "迭代次数：7740,  loss值：0.95033073\n",
      "迭代次数：7750,  loss值：0.95033073\n",
      "迭代次数：7760,  loss值：0.95033085\n",
      "迭代次数：7770,  loss值：0.95033073\n",
      "迭代次数：7780,  loss值：0.95033079\n",
      "迭代次数：7790,  loss值：0.95033079\n",
      "迭代次数：7800,  loss值：0.95033073\n",
      "迭代次数：7810,  loss值：0.95033073\n",
      "迭代次数：7820,  loss值：0.95033097\n",
      "迭代次数：7830,  loss值：0.95033097\n",
      "迭代次数：7840,  loss值：0.95033097\n",
      "迭代次数：7850,  loss值：0.95033097\n",
      "迭代次数：7860,  loss值：0.95033103\n",
      "迭代次数：7870,  loss值：0.95033091\n",
      "迭代次数：7880,  loss值：0.95033097\n",
      "迭代次数：7890,  loss值：0.95033091\n",
      "迭代次数：7900,  loss值：0.95033085\n",
      "迭代次数：7910,  loss值：0.95033091\n",
      "迭代次数：7920,  loss值：0.95033091\n",
      "迭代次数：7930,  loss值：0.95033091\n",
      "迭代次数：7940,  loss值：0.95033085\n",
      "迭代次数：7950,  loss值：0.95033085\n",
      "迭代次数：7960,  loss值：0.95033091\n",
      "迭代次数：7970,  loss值：0.95033091\n",
      "迭代次数：7980,  loss值：0.95033085\n",
      "迭代次数：7990,  loss值：0.95033085\n",
      "迭代次数：8000,  loss值：0.95033091\n",
      "迭代次数：8010,  loss值：0.95033091\n",
      "迭代次数：8020,  loss值：0.95033097\n",
      "迭代次数：8030,  loss值：0.95033073\n",
      "迭代次数：8040,  loss值：0.95033073\n",
      "迭代次数：8050,  loss值：0.95033103\n",
      "迭代次数：8060,  loss值：0.95033109\n",
      "迭代次数：8070,  loss值：0.95033097\n",
      "迭代次数：8080,  loss值：0.95033097\n",
      "迭代次数：8090,  loss值：0.95033103\n",
      "迭代次数：8100,  loss值：0.95033103\n",
      "迭代次数：8110,  loss值：0.95033073\n",
      "迭代次数：8120,  loss值：0.95033079\n",
      "迭代次数：8130,  loss值：0.95033067\n",
      "迭代次数：8140,  loss值：0.95033073\n",
      "迭代次数：8150,  loss值：0.95033067\n",
      "迭代次数：8160,  loss值：0.95033067\n",
      "迭代次数：8170,  loss值：0.95033067\n",
      "迭代次数：8180,  loss值：0.95033073\n",
      "迭代次数：8190,  loss值：0.95033115\n",
      "迭代次数：8200,  loss值：0.95033103\n",
      "迭代次数：8210,  loss值：0.95033109\n",
      "迭代次数：8220,  loss值：0.95033103\n",
      "迭代次数：8230,  loss值：0.95033103\n",
      "迭代次数：8240,  loss值：0.95033097\n",
      "迭代次数：8250,  loss值：0.95033067\n",
      "迭代次数：8260,  loss值：0.95033073\n",
      "迭代次数：8270,  loss值：0.95033073\n",
      "迭代次数：8280,  loss值：0.95033062\n",
      "迭代次数：8290,  loss值：0.95033067\n",
      "迭代次数：8300,  loss值：0.95033067\n",
      "迭代次数：8310,  loss值：0.95033073\n",
      "迭代次数：8320,  loss值：0.95033091\n",
      "迭代次数：8330,  loss值：0.95033091\n",
      "迭代次数：8340,  loss值：0.95033091\n",
      "迭代次数：8350,  loss值：0.95033085\n",
      "迭代次数：8360,  loss值：0.95033121\n",
      "迭代次数：8370,  loss值：0.95033121\n",
      "迭代次数：8380,  loss值：0.95033121\n",
      "迭代次数：8390,  loss值：0.95033121\n",
      "迭代次数：8400,  loss值：0.95033127\n",
      "迭代次数：8410,  loss值：0.95033115\n",
      "迭代次数：8420,  loss值：0.95033115\n",
      "迭代次数：8430,  loss值：0.95033121\n",
      "迭代次数：8440,  loss值：0.95033085\n",
      "迭代次数：8450,  loss值：0.95033085\n",
      "迭代次数：8460,  loss值：0.95033085\n",
      "迭代次数：8470,  loss值：0.95033079\n",
      "迭代次数：8480,  loss值：0.95033079\n",
      "迭代次数：8490,  loss值：0.95033079\n",
      "迭代次数：8500,  loss值：0.95033079\n",
      "迭代次数：8510,  loss值：0.95033079\n",
      "迭代次数：8520,  loss值：0.95033079\n",
      "迭代次数：8530,  loss值：0.95033085\n",
      "迭代次数：8540,  loss值：0.95033073\n",
      "迭代次数：8550,  loss值：0.95033073\n",
      "迭代次数：8560,  loss值：0.95033079\n",
      "迭代次数：8570,  loss值：0.95033103\n",
      "迭代次数：8580,  loss值：0.95033109\n",
      "迭代次数：8590,  loss值：0.95033109\n",
      "迭代次数：8600,  loss值：0.95033097\n",
      "迭代次数：8610,  loss值：0.95033103\n",
      "迭代次数：8620,  loss值：0.95033103\n",
      "迭代次数：8630,  loss值：0.95033103\n",
      "迭代次数：8640,  loss值：0.95033103\n",
      "迭代次数：8650,  loss值：0.95033103\n",
      "迭代次数：8660,  loss值：0.95033109\n",
      "迭代次数：8670,  loss值：0.95033097\n",
      "迭代次数：8680,  loss值：0.95033097\n",
      "迭代次数：8690,  loss值：0.95033103\n",
      "迭代次数：8700,  loss值：0.95033091\n",
      "迭代次数：8710,  loss值：0.95033097\n",
      "迭代次数：8720,  loss值：0.95033085\n",
      "迭代次数：8730,  loss值：0.95033073\n",
      "迭代次数：8740,  loss值：0.95033073\n",
      "迭代次数：8750,  loss值：0.95033079\n",
      "迭代次数：8760,  loss值：0.95033079\n",
      "迭代次数：8770,  loss值：0.95033079\n",
      "迭代次数：8780,  loss值：0.95033079\n",
      "迭代次数：8790,  loss值：0.95033067\n",
      "迭代次数：8800,  loss值：0.95033091\n",
      "迭代次数：8810,  loss值：0.95033091\n",
      "迭代次数：8820,  loss值：0.95033097\n",
      "迭代次数：8830,  loss值：0.95033115\n",
      "迭代次数：8840,  loss值：0.95033115\n",
      "迭代次数：8850,  loss值：0.95033121\n",
      "迭代次数：8860,  loss值：0.95033109\n",
      "迭代次数：8870,  loss值：0.95033109\n",
      "迭代次数：8880,  loss值：0.95033109\n",
      "迭代次数：8890,  loss值：0.95033109\n",
      "迭代次数：8900,  loss值：0.95033109\n",
      "迭代次数：8910,  loss值：0.95033109\n",
      "迭代次数：8920,  loss值：0.95033109\n",
      "迭代次数：8930,  loss值：0.95033115\n",
      "迭代次数：8940,  loss值：0.95033115\n",
      "迭代次数：8950,  loss值：0.95033115\n",
      "迭代次数：8960,  loss值：0.95033115\n",
      "迭代次数：8970,  loss值：0.95033115\n",
      "迭代次数：8980,  loss值：0.95033103\n",
      "迭代次数：8990,  loss值：0.95033103\n",
      "迭代次数：9000,  loss值：0.95033109\n",
      "迭代次数：9010,  loss值：0.95033067\n",
      "迭代次数：9020,  loss值：0.95033073\n",
      "迭代次数：9030,  loss值：0.95033073\n",
      "迭代次数：9040,  loss值：0.95033115\n",
      "迭代次数：9050,  loss值：0.95033121\n",
      "迭代次数：9060,  loss值：0.95033127\n",
      "迭代次数：9070,  loss值：0.95033121\n",
      "迭代次数：9080,  loss值：0.95033127\n",
      "迭代次数：9090,  loss值：0.95033127\n",
      "迭代次数：9100,  loss值：0.95033127\n",
      "迭代次数：9110,  loss值：0.95033115\n",
      "迭代次数：9120,  loss值：0.95033115\n",
      "迭代次数：9130,  loss值：0.95033121\n",
      "迭代次数：9140,  loss值：0.95033115\n",
      "迭代次数：9150,  loss值：0.95033121\n",
      "迭代次数：9160,  loss值：0.95033121\n",
      "迭代次数：9170,  loss值：0.95033121\n",
      "迭代次数：9180,  loss值：0.95033121\n",
      "迭代次数：9190,  loss值：0.95033121\n",
      "迭代次数：9200,  loss值：0.95033121\n",
      "迭代次数：9210,  loss值：0.95033121\n",
      "迭代次数：9220,  loss值：0.95033127\n",
      "迭代次数：9230,  loss值：0.95033121\n",
      "迭代次数：9240,  loss值：0.95033091\n",
      "迭代次数：9250,  loss值：0.95033091\n",
      "迭代次数：9260,  loss值：0.95033091\n",
      "迭代次数：9270,  loss值：0.95033091\n",
      "迭代次数：9280,  loss值：0.95033097\n",
      "迭代次数：9290,  loss值：0.95033097\n",
      "迭代次数：9300,  loss值：0.95033103\n",
      "迭代次数：9310,  loss值：0.95033103\n",
      "迭代次数：9320,  loss值：0.95033103\n",
      "迭代次数：9330,  loss值：0.95033109\n",
      "迭代次数：9340,  loss值：0.95033103\n",
      "迭代次数：9350,  loss值：0.95033109\n",
      "迭代次数：9360,  loss值：0.95033109\n",
      "迭代次数：9370,  loss值：0.95033097\n",
      "迭代次数：9380,  loss值：0.95033097\n",
      "迭代次数：9390,  loss值：0.95033097\n",
      "迭代次数：9400,  loss值：0.95033097\n",
      "迭代次数：9410,  loss值：0.95033097\n",
      "迭代次数：9420,  loss值：0.95033103\n",
      "迭代次数：9430,  loss值：0.95033097\n",
      "迭代次数：9440,  loss值：0.95033103\n",
      "迭代次数：9450,  loss值：0.95033103\n",
      "迭代次数：9460,  loss值：0.95033103\n",
      "迭代次数：9470,  loss值：0.95033103\n",
      "迭代次数：9480,  loss值：0.95033103\n",
      "迭代次数：9490,  loss值：0.95033091\n",
      "迭代次数：9500,  loss值：0.95033091\n",
      "迭代次数：9510,  loss值：0.95033091\n",
      "迭代次数：9520,  loss值：0.95033091\n",
      "迭代次数：9530,  loss值：0.95033091\n",
      "迭代次数：9540,  loss值：0.95033091\n",
      "迭代次数：9550,  loss值：0.95033097\n",
      "迭代次数：9560,  loss值：0.95033079\n",
      "迭代次数：9570,  loss值：0.95033079\n",
      "迭代次数：9580,  loss值：0.95033079\n",
      "迭代次数：9590,  loss值：0.95033079\n",
      "迭代次数：9600,  loss值：0.95033085\n",
      "迭代次数：9610,  loss值：0.95033085\n",
      "迭代次数：9620,  loss值：0.95033073\n",
      "迭代次数：9630,  loss值：0.95033073\n",
      "迭代次数：9640,  loss值：0.95033073\n",
      "迭代次数：9650,  loss值：0.95033079\n",
      "迭代次数：9660,  loss值：0.95033073\n",
      "迭代次数：9670,  loss值：0.95033073\n",
      "迭代次数：9680,  loss值：0.95033079\n",
      "迭代次数：9690,  loss值：0.95033079\n",
      "迭代次数：9700,  loss值：0.95033109\n",
      "迭代次数：9710,  loss值：0.95033115\n",
      "迭代次数：9720,  loss值：0.95033109\n",
      "迭代次数：9730,  loss值：0.95033115\n",
      "迭代次数：9740,  loss值：0.95033115\n",
      "迭代次数：9750,  loss值：0.95033103\n",
      "迭代次数：9760,  loss值：0.95033103\n",
      "迭代次数：9770,  loss值：0.95033103\n",
      "迭代次数：9780,  loss值：0.95033109\n",
      "迭代次数：9790,  loss值：0.95033103\n",
      "迭代次数：9800,  loss值：0.95033109\n",
      "迭代次数：9810,  loss值：0.95033103\n",
      "迭代次数：9820,  loss值：0.95033103\n",
      "迭代次数：9830,  loss值：0.95033103\n",
      "迭代次数：9840,  loss值：0.95033109\n",
      "迭代次数：9850,  loss值：0.95033109\n",
      "迭代次数：9860,  loss值：0.95033109\n",
      "迭代次数：9870,  loss值：0.95033109\n",
      "迭代次数：9880,  loss值：0.95033091\n",
      "迭代次数：9890,  loss值：0.95033091\n",
      "迭代次数：9900,  loss值：0.95033091\n",
      "迭代次数：9910,  loss值：0.95033097\n",
      "迭代次数：9920,  loss值：0.95033091\n",
      "迭代次数：9930,  loss值：0.95033097\n",
      "迭代次数：9940,  loss值：0.95033097\n",
      "迭代次数：9950,  loss值：0.95033097\n",
      "迭代次数：9960,  loss值：0.95033097\n",
      "迭代次数：9970,  loss值：0.95033097\n",
      "迭代次数：9980,  loss值：0.95033097\n",
      "迭代次数：9990,  loss值：0.95033097\n",
      "迭代次数：10000,  loss值：0.95033091\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXAAAAD4CAYAAAD1jb0+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nO3dd3RU1d7G8e+emfSEFNKAJJTQCT10pFdRQLArIHJFsaLY9drFdhUVFcWKiCgqRUBpGpAOAUJvoSa0hHSSkGRm9vtHcu+LGCWQTE5m8vusxUpmMjPnOUEfzuzZZx+ltUYIIYTzMRkdQAghxJWRAhdCCCclBS6EEE5KClwIIZyUFLgQQjgpS2VuLDg4WNerV68yNymEEE5vy5YtZ7XWIRffX6kFXq9ePeLj4ytzk0II4fSUUsdKu1+GUIQQwklJgQshhJOSAhdCCCclBS6EEE5KClwIIZyUFLgQQjgpKXAhhHBSUuBCCOFIWcmw/Dk4l1LhLy0FLoQQjhT/Jax9H4ryK/ylpcCFEMJRrAWw5StoMhgC61b4y0uBCyGEo+yeD3lnOVz/Voe8vBS4EEI4gNVmJ2PlVJJMEfSdD7tPZlX4NqTAhRCigu06kcWEN6cTmLGT+W6DeX1kaxqG+lb4dip1NUIhhHB1drvmmXk7uavwV6wWH+6b+G9MXv4O2ZYcgQshRAVatPMUSclJDFZrsbS9xWHlDVLgQghRYQqsNt5cso8HA9ZhthdBh7scuj0pcCGEqCBfrzvG6YwcbjEtgwa9ILSpQ7cnBS6EEBUgM6+Qqb8fZGLEATzyTkOnexy+TSlwIYSoALM2HienwMqdlqUQWA8aDXD4NqXAhRCiAsQfTefqmqfxPr0ZOo4Hk9nh25QCF0KIctJak5CUyVjLMnDzgba3V8p2ZR64EEKU0/H0PEx5Z2mrf4PYO8DTcVMHL1SmAldKHQVyABtg1VrHKqWCgO+BesBR4EatdYZjYgohRNWVkJTJLebfMeui4uGTSnI5Qyi9tdZttNaxJbefBH7TWjcCfiu5LYQQ1c7OYymMtizHHt0XQhpX2nbLMwY+DJhR8v0MYHj54wghhPPxTVxIqMrE1OXeSt1uWQtcA8uUUluUUv99fxCmtT4FUPI1tLQnKqXGK6XilVLxqamp5U8shBBVSGGRjf7ZP5HqWQ+i+1bqtsta4N201u2AwcB9SqkeZd2A1nq61jpWax0bEhJyRSGFEKKqSkpYQQt1lFPNxoJSlbrtMhW41vpkydcUYB7QETijlKoFUPK14i/4JoQQVZxl88dkaF9qdqmcqYMXumSBK6V8lFJ+//0eGADsAn4GxpQ8bAywwFEhhRCiSko/QmRKHPNMA6gdUrPSN1+WaYRhwDxV/NbAAnyrtV6ilNoMzFFKjQOOAzc4LqYQQlRBm6Zjx8SeiBtRlTx8AmUocK31YaB1KfenAZU7Yi+EEFVFfiZ6ywwW2jpTv0EjQyLIqfRCCHEltn6NKsrlM+vVtI4IMCSCFLgQQlwuWxFs/JikGu3ZQ31aRVbOqfMXkwIXQojLtWcBZJ/g7XP96dKgJjU83QyJIYtZCSHE5dAa1k0l3asuCzJimDuwiWFR5AhcCCEux7F1cCqBqXkDGNCiFm2jAg2LIkfgQghxOdZNJc/sz3d53Vho4NE3yBG4EEKUXep+OPArnxf2Y2j7aBqG+hkaR47AhRCijPS6DyhS7nyjBzKvnzFzvy8kR+BCCFEWOWewb5/N90U9GN23PbUDvIxOJAUuhBBlkRY3FWWzkhBxGxN6RhsdB5ACF0KIS8o/l4Xbti+JM3XiiduuxmSq/HVPSiNj4EIIUYrElHOs3J/CwTPnqJf4NRP0OWoOfJRQP0+jo/2PFLgQQpTigdnb2HsqmzBvE4vUAlIC29GmS3+jY/2JDKEIIcRFzhfZOHAmh3t7RbNxWBYhthRCB1e967ZLgQshxEX2nc7BZte0quMHa96F0ObQaIDRsf5CClwIIS6y+2QWAO0L4yF1L3R7qNKvd1kWUuBCCHGRXSeyqeFpIXj7NPCPhJiRRkcqlRS4EEJcZM/JLK6rmYw6vh663A9mY5aLvRQpcCGEuECRzc7e0zncZp0LXkHQbpTRkf6WFLgQQlzgUOo5GtiO0jhrDXS6G9x9jI70t6TAhRDiArtPZDPB8jN2Nx/oON7oOP9IClwIIS5w8vBurjGth9hx4B1kdJx/5BwFvuJFWDjR6BRCiGqg2eEvsSkLpq73GR3lkpyjwG2FsPVrSD9idBIhhAuzZ56gZ94yttUcAn7hRse5JKco8H31R6OVGda+a3QUIYQLO7fyXRSaMy3vNjpKmThFgT/3ezrfFPWgaMss1ifswG7XRkcSQria3LN475zJz/au1G/Ywug0ZVLmAldKmZVS25RSi0pu11dKbVRKHVRKfa+UcndUyHdvbkN+7P0o7Oz98VXGfrXZUZsSQlRX6z/EbDvPx7bhNA73NTpNmVzOEfhDwN4Lbr8BTNFaNwIygHEVGexCtQO8GD+sN6rVTYxyj2P3gUQSU845anNCiOomPwM2fcom7x6YQpvgYTEbnahMylTgSqkIYAjwWcltBfQBfix5yAxguCMCXsjcYxIWeyHjLL+wcPtJR29OCFFdbPwECnN4/dwQ2kYFGJ2mzMp6BP4u8DhgL7ldE8jUWltLbicDdSo4218FN0TFjOAOtxWs2r4PrWUsXAhRTuezYcM0DtfsybbCCG7vXNfoRGV2yQJXSl0DpGitt1x4dykPLbVNlVLjlVLxSqn41NTUK4x5gR6P4aXz6ZP5E7tPZpf/9YQQ1dvmz+B8Js9lXk3PxiG0qO1vdKIyK8sReDdgqFLqKPAdxUMn7wIBSqn/XpItAih1TENrPV1rHau1jg0JCSl/4tBmFDa+lrHmpSzfuq/8ryeEqL4Kc2H9B5wI7saa3Egm9KoaV5svq0sWuNb6Ka11hNa6HnAz8LvW+jYgDri+5GFjgAUOS3kR9z5P4qfyqZHwuQyjCCGu3ObPIC+Nl3OuoW1UAJ3qV+1T5y9WnnngTwCPKKUSKR4T/7xiIpVBeAwnwvtxg3Uh2w8eq7TNCiFcSGEurH2flNDuLMmqyz09o1FV8Ko7/+SyClxrvVJrfU3J94e11h211g211jdorQscE7F0AYOepobKIz3ug8rcrBDCyWmtee3XvXw/7QXIO8tjqYOIDvGhf7Mwo6NdNqc4E7M0PvXas8O7C7GnvsWel2l0HCGEk9h5IouvV+1hQOb3bHdvh6VuZ14aFoPJ5FxH3+DEBQ6Q1uFhapBLyor3jI4ihHASCxJOMsbtdwJ1Fq1vf53P7+hAt4bBRse6Ik5d4K079maZrT0B26dDvhyFCyH+mc2uWZ5wmHvdF0GD3hDVyehI5eLUBR7k487SkLF42s7BhmlGxxFCVHEbD6cxOH8RNWyZ0Ospo+OUm1MXOECDll341dYB+/oPi9czEEKIv/HrlkTusSzC1qCv0x99gwsUeJ+mobxnHYmpMAfWf2h0HCFEFXW+yEbw3i8JVDmY+zxjdJwK4fQF3jTcj6wajdni06N4GCU3zehIQogqaM2uw4zRC0mr0wci2hsdp0I4fYErpejdNJQXc4ahC3NhncxIEUL81fnVUwlQufgPft7oKBXG6QscoG/TUHYU1iK13jDYOJ0NO/bw2q97KbLZL/1kIYTLy8lIoWfaHPYE9MIS0cboOBXGJQq8a3QwHhYTX7ndhM1ayN45L/DJqsP8vi/F6GhCiCrg1C9v4MN5dM8njI5SoVyiwL3czXSJrslHOzVzbD0Y5fY7rfyymbM5yehoQgij5ZyhXuJMlpm706x1F6PTVCiXKHCA0V3q0r5uIC1uehmLSfFy4K/E7U/hdNZ5o6MJIQxUuPItTPYiDjR9wClPl/8nLlPgfZqG8dOErrSKaQntx9IqdRFRnOLHLXIULkS1lXkcy9avmGPrSecOHYxOU+FcpsD/pMejKIsHkwMWMic+Gbtd1gwXolpa9QY24FvPm2lfN9DoNBXONQvcNxQ630vX/JX4Zexhw2GZGy5EtXP2IDrhW2bZ+tOuZQxmFxs+AVctcIBuD6K9AnnS4we+kw8zhah+fnsJm9mTqYXXMjimltFpHMJ1C9zTH9X9Ya5iG2d3x5GVV2R0IiFEZUneAnt/ZmmNG1G+IXR0skullZXrFjhAx/EUeYfxiGk2y3afMjqNEKIyaA0rnsfuHcxzqb0Y2CLcJYdPwNUL3M0LS5+niTUd4NTmuUanEUJUhkO/w9HVfMoIsu0e3BgbaXQih3HtAgdU29s56xnFkDOfkJ2Xb3QcIYQj2e0ULXue0yqUqdlXMX1ULK0jA4xO5TAuX+CYLWR1fZpodZJDSz8xOo0QwoHObf0Bt5SdTLHdwPSxXendNNToSA7l+gUO1O92E9tVE+rveh8Kc42OI4RwBGsBtuXPs8del+vveJiu0c55ncvLUS0K3GQ2sbnhRAJsaRSu+cDoOEIIByjcMB3/glMsrX0vHRqEGB2nUlSLAgeI6TKQZbb2qHXvQe5Zo+MIISpSfgZ65Zv8YWvJVYNuNDpNpak2Bd6hXhDT3UdhsubDqjeMjiOEqED21VNws+YwL3i8S54y/3eqTYGbTYrGMbH8oPui47+AsweNjiSEqAiZSegN05hn687Avv1RyjXnfJem2hQ4wNUxtXirYARWkwcsd53LKglRnenfXsJqh+99R9O/ebjRcSrVJQtcKeWplNqklNqulNqtlHqx5P76SqmNSqmDSqnvlVLujo9bPl2ja+IVGM5c7xtg/2I4usboSEKI8kjegto5h0+tgxnas5PLnnH5d8pyBF4A9NFatwbaAIOUUp2BN4ApWutGQAYwznExK4bJpLilYxTPpfTC6lsLlj4DdrluphBOSWuyFjzOWe3P2vBRXN8+wuhEle6SBa6LnSu56VbyRwN9gB9L7p8BDHdIwgp2Q/sIbCYPfgkdD6cSYOccoyMJIa7AoT++xT81nm+8R/Hxnb3wdDMbHanSlWkMXCllVkolACnAcuAQkKm1tpY8JBmo8zfPHa+UildKxaemplZE5nIJreFJv2ZhvHi0BfZabWHFi3JyjxBOJPt8Ed+tS8Q97kUOmepy2z3P4O/tZnQsQ5SpwLXWNq11GyAC6Ag0K+1hf/Pc6VrrWK11bEhI1Zhcf2unKNLyrKxtOAlyTsLa942OJIS4hLPnCpj43TY6vrqCQ4vfJpIz+A97gxB/b6OjGeayZqForTOBlUBnIEApZSn5UQRwsmKjOU73hsFEBXnzQWIwtBgBa9+DrGSjYwkh/sFHcYdYvPMUo1t686T3QnSjgQS3Hmx0LEOVZRZKiFIqoOR7L6AfsBeIA64vedgYYIGjQlY0k0lxc8dINh5J50jbx0HbYcULRscSQvyDVQdS6BIdzNOeP2K25aMGTjY6kuHKcgReC4hTSu0ANgPLtdaLgCeAR5RSiUBN4HPHxax4N8ZG4uth4aXVOeiuD8DOHyBpk9GxhBClOJGZz6HUXK4LPwtbZ0KneyC4odGxDFeWWSg7tNZttdattNYxWuuXSu4/rLXuqLVuqLW+QWtd4Pi4FSfY14OJ/RoRtz+VuODbwK8W/Pq4TCsUogr640AqoBmU9C5414QejxkdqUqoVmdiXmxM13o0DvPluSVHKez9PJzcBgmzjI4lhLjIqv2p3O67Fa9TG6HPs+DluhdpuBzVusDdzCZeHBpDckY+H55tBxEdi8fC8zONjiaEKGG12dl6KJlH1UwIbwntRhsdqcqo1gUO0CW6Jte2rs20Pw4zN/xBdF4amUteMTqWEKJEQlImo6xzCShKgav/A6bqd8LO36n2BQ7wzNXNqO3vySNrTMy29sIn4Qu+mPeL0bGEEMCO7VsZb15EYfMbIKqz0XGqFMulH+L6wv09WflYbzJyC0lKbkLR9z1pvPUV1rToQPfGVePkIyGqq5hdb2A3ueE1WN4ZX0yOwC8Q6ONOqyYNsfR/nu7m3fzy/Uek5xYaHUuIaitn52I6Fm0ivu548KteS8WWhRR4Kdw7jSO/ZgwPWb/kuR82oHWpqwQIIRypKB/Tr0+QaK+NX8/7jE5TJUmBl8Zkxmv4u4SpDGISP+G5BbvJyisyOpUQ1UrSwsn45CXxnsc9tKwbanScKkkK/O9EdkC3HcVdll/ZsGktPd6K47PVhym0yok+Qjja0lVrCd0+jd8sPXl8wl3V7kINZSUF/g9Uvxcwe/oxv+5cWtWpwSuL9/LInASjYwnh0r5ZfxTPFU9iM7nR4e4PiQyqvqsNXooU+D/xCYb+L+JzagMzYw/zUN9GLNpxipX7U4xOJoRLOl9kY/vymfQ078BjwHPUCIk0OlKVJgV+KW1HQ2QnWPYs93YKpEGID88t2M35IpvRyYRwOQs37WeS7XPOBTbH3PEuo+NUeVLgl2IywTVTID8Tj7gXeWVYDMfT8/gwLtHoZEK4FLtdY175CqEqE5+RU8Esp6lcihR4WYS1gC73wbaZdHU7yPA2tfl41SEOpZ679HOFEGWyed0Khhf+wtEGt6IiYo2O4xSkwMuq15PgHwWLJvLMwGg83cw8t2CXzBEXoiLYrISufII0FUjUyFeNTuM0pMDLyt0HhrwNqfsI2T6NRwc0YW1iGkt2nTY6mRBO78SSd6hvPcSOlk9j8Qk0Oo7TkAK/HI0HQMxIWP0fbmuQT9NwP15ZvJf8QvlAU4jLZbdrNh9N570fllFz81usoh2dhtxhdCynIgV+uQa9Dm7eWBZP5IVrm3EiM59pqw4ZnUoIpzPx+wRu+HgdsTtfQisL/tdPxdfTzehYTkUK/HL5hsLAVyFpA53Tf+ba1sUfaCal5xmdTAinkZpTwKIdJ3mz4W66mXbiNfgl2sTEGB3L6UiBX4k2t0H9HrD8eZ69yg+zUry5dL/RqYRwGot3nCRIZzEy9UOI6gKx44yO5JSkwK+EUnDt+6BthK16iptiI1i2+zS5BVajkwnhFBZsP8nbft9ituYX/79kkiq6EvJbu1JB9aHvc3BwGbd7b6DAamfl/lSjUwlR5R1PyyMseRk9i1ZDz8chpLHRkZyWFHh5dBwPkZ2I3vIyjb1zWbJbphQKcSnL4nfzstsXFIa2gm4TjY7j1KTAy8NkhqEfoIryecd3Jr/vPS1rpAjxD7TW1N/8MoEqD/cR08Ass07KQwq8vEIaQ++nicn+g77W1axNPGt0IiGqrOT1P9DXuordDcdDuMw6KS8p8IrQ9QHsER142f0r1m3bZXQaIaqm3DQC455gj70ukUOfNTqNS7hkgSulIpVScUqpvUqp3Uqph0ruD1JKLVdKHSz5Wn3PfzWZMQ3/GC9VRM8Dr1BklWEUIf5Ea6wLH8ajKJvvI54mqIaP0YlcQlmOwK3AJK11M6AzcJ9SqjnwJPCb1roR8FvJ7eoruCGHWk2iB1s5+tunRqcRwjBaa+6eGc/8bSf+/85dP2HZt4ApRSO4dsAA48K5mEsWuNb6lNZ6a8n3OcBeoA4wDJhR8rAZwHBHhXQW9a6exGbdjMiNL0LGMaPjCGGIgynnWLr7DE/N3cnRs7mQfQq9eBLbacS+6DuJrRdkdESXcVlj4EqpekBbYCMQprU+BcUlD5R62Wil1HilVLxSKj411bXnSXt5uPFL9HMU2TQnvrwDm1VO7BHVz38/yFcKHp2TgP75AayF+TxUcA8PD2hucDrXUuYCV0r5Aj8BE7XW2WV9ntZ6utY6VmsdGxISciUZncqkmwawoNZD1MneypypT5CaU2B0JCEq1drENKKCvHlleAxNT/yASlzOm7bbaNq8LS0j/I2O51LKVOBKKTeKy3uW1npuyd1nlFK1Sn5eC5Ar/QK+HhZuG/8ESWH9GJn5JU9P+1Yu+iCqDavNzsbDaXRrGMx1Eed4zn0WK22t+aywL48MkDMuK1pZZqEo4HNgr9b6nQt+9DMwpuT7McCCio/nnJTJROTo6dg8A5l07m12HTtjdCQhKsXOE1nkFFjp3sAPNW88Fk8/3vB8gBvbR9E4zM/oeC6nLEfg3YBRQB+lVELJn6uB14H+SqmDQP+S2+K/fGpivfYDmpqSKFzyb6PTCFEp/jv+3efk53BqO6ZhU5n72AheG9HS4GSu6ZKXfdZarwHU3/y4b8XGcS1+MYP55dcRXH16DuwfCU0GGR1JCIdam5jGzcFH8No0FdqNgaZD8DI6lAuTMzEdLL3rU+yx18U2bwLkyGJXwnWdL7Jx+PgxnimYAsGNYNBrRkdyeVLgDtYvJooHiu7HXpgL8+4Gu93oSEI4RPyRdCarafjYsuD6L4ovBC4cSgrcwcL9PfGLaMHHXuPh8EpYO8XoSEI4xPm1H9HXvI2ivi9BuIx5VwYp8EowKCact9M6k9d4OPz+Khxbb3QkISpModVO1qHN9Dz2PvEeHfHoOsHoSNWGFHglGNgiHFD8VPtRCIiCn8ZBXrrRsYQol7RzBXR8dQXtn/2JzBm3kqprsKXN5OJTMEWlkAKvBPWDfWga7sdPu7PJH/455KbC/AkyHi6c2oq9Z0jJOc8PtWYRaU7j4FXvcVuftkbHqlakwCvJ9e0jSEjKpN1nqfwYfA8cWELigleJ25/C9qRMo+MJcdmW7T7DQ75xNM2Iw9TveXr2G4qvxyVnJosKJL/tSjKue33aRAYwd9sJXtreDU/bJgYnvM2zmz3ZYG/Owvu7yzoRwmnkFVrJTlzPg25fQeNB0OUBoyNVS3IEXkmUUsTWC2LydS3Z/Gx/6o/9gqKABnxd42Oi3LKZtVGWnxXOY/3OA7xnnkKRdzgMnwYmqRIjyG/dAB4WMy0a1MHztm9xt+Xzld80FiccJ/t8kdHRhLg0u43I3++npsrBcss34C3rextFCtxIoU1h6Ps0yNvOw3omCy68gokQVZT9t5dpnLuFubUexhIhH1oaSQrcaC2vR3e+lzstSzi9+itZelZUOfmFNn6ITyK3wAp7F2JaO4Vvrb3x73qn0dGqPSnwKkD1f5kzQR14IPcD9m5dY3QcIf7k2fm7eOzHHTw8dTb2ueM54dOcyYylR2PXv0BLVScFXhWYLfiM+oYMahC+ZBzknjU6kRAAzNuWzE9bk7kpxpd/n3uZtCJ37syfSIfoWjJlsAqQAq8ifAPDmdv4DbwL0yn69lawFhodSVRzR87m8uy8XXSu689r9ilEmDJ43e8Z9uf50r95uNHxBFLgVcqAfoN4wnYPbic2wuKHQcbDhUEKrDYemL0Vi9nEZ7XmYzoch7rmbV59cBzv39KWG2IjjI4okAKvUhqF+RHa5Vbetw6Hbd/AhmlGRxLV1PxtJ9h1Iptv2+7BN+FT6HwvtBuNp5uZoa1r42aW6qgK5G+hipnYrzFzfEaxxtIZvewZOLDM6EiiGlq+J4Whfgdpvu0laDQABrxidCRRCinwKsbHw8ILw1oy/tx4zvo0wv7DHXzx489c/d5qdp3IMjqeqAbOF9k4kbiDN+z/QQU3hpGfg8lsdCxRCinwKqhf8zB6tKjH0LQHOFPoweCdD5Fx+ijz5UQfUQni9xzgEzUZs8UNbv0OPGsYHUn8DSnwKurFYS1o2awZazt+RJh7Ad/6TmFLYrLRsYSrK8yj3tI7CVFZ6Fu+g8B6RicS/0AKvIoKq+HJ9NGxXD9kMKYbZ1C36AgPpb1CWtY5o6MJV2W3oef+i9p5e/k8/Fk86nUyOpG4BClwZ9CoP0ndJtPLvJ28n+6T6YWi4mkNS55E7VvMi0Wjqdn+OqMTiTKQAncSdfrczYf6BiKPz4ffX/7f/TuTs0g7V2BgMuES/vgPbJrOtshRzLANpHeTUKMTiTKQAncSFrOJbfXH87O5P6x+GzZ+wqHUcwz7cA0DpvzB8j1njI4onNWWryDuFWh9C5MLbiamTg3C/T2NTiXKQArciXRtGMLDuaPJazAIfn2c9fM+wt1iIrSGJ3d9Hc+TP+0oXjFOiH+QmlPAzuSSKal7F8Kih6Fhf9L7vs2WpCz6NA0zNqAos0uuRqOU+gK4BkjRWseU3BcEfA/UA44CN2qtMxwXUwB0bxSMDTOLG7/CNQU53Jz8Gh5NX2fYTXcxZcUBPl51CIDXR7YyOKmoqjJyC7nh43WcyMxn9QhN+OKxUKc93DiDpQlp2DX0bSrDJ86iLEfgXwGDLrrvSeA3rXUj4LeS28LBGoX6EuLnweoj53jJ51n2UJ/rD/8b96Q1PDGoKTd3iGRBwkly5Mo+ohSFVjsTZm3hZOZ5OpoOELRoLIQ0hdt+IMvmztvLDtAqwp+WdeTarM7ikgWutf4DSL/o7mHAjJLvZwDDKziXKIVSiq7RNYnbn8L3OzNY3vZDVFAD+PYmOLaemzpEkV9k4+ftJ42OKqoYrTXP/7yLDYfT+biviS/c3yTJFsTmqz4Hr0DeWrqP9NwCJl/XEpNJGR1XlNGVjoGHaa1PAZR8lfdclaRbdDA55624W0yM6tsWRi+AGrVh1g205iBNw/34fnOS0TFFFTNr43Fmb0rihQ5W+mwaj7tvEJM8X+SF31PYciyDWRuPM6ZrPWLk6NupOPxDTKXUeKVUvFIqPjU11dGbc3ndGgWjFNzWqS6hfp7gFwZjFoJPMOqbkdzXJIcdyVnsPinrpohiWmu+WHOEEbUzGJP4EHj4oe5YxNiru7P7ZDZ3fLmJMD9PJg1oYnRUcZmutMDPKKVqAZR8Tfm7B2qtp2utY7XWsSEhcgmm8qoT4MWC+7rx+KAL/merUbu4xL38GZIwgfaWI/87Crfa7ExbeUjWUanGDpw5h1vaXibnPINy8y7+byWwHkNb16ZNZAA55628MLS5XGHHCV3p39jPwBjg9ZKvCyoskbikVhEBf70zIBLuWIxpxrV8UzSZu7Zpjnarz+M/7mDT0XR83M30bhqKv5db5QcWhopfH8ds91dw8/ArLu+g+kDxZypTb2nLhsNpDGwhV9hxRpc8AldKzQbWA02UUslKqXEUF3d/pdRBoH/JbWG0gCi4YzH4BPOxfoWn3p3OzhNZPNinIbmFNmZvOm50QlHZkuMZtv0ebGZvzHf+AjWj//TjyCBvboiNRCn54NIZlWUWyi1a61paazetdYTW+nOtdZrWuq/WulHJ16e+dLIAABGVSURBVItnqQij+Efg+a9fyTAH86VlMr8NK+KRAU3o1rAmX649QqHVbnRCUVmOrcM+Yyhn7b7Edf0KghoYnUhUMDkT0wUp/zqEPLACj7DG1P5lLOyez/ge0ZzJLmBBgoyFVwsHlsLM68h0C+Gmwn9zVYd2RicSDiAF7qI8A2uh7lhcfJbdj2PpkfMLTcP9+HT1YbSsZujatn8Hs2+BkKbc7/EqtaMaUMvfy+hUwgGkwF2ZVwCMmgsNeqMWPsg7YUs5cCaHlQdkOqfLWv8RzLsb6nUjedgPrDulGBwjH1C6Kpk35OrcfeCW7+DnB2i+4wOmeB/gP7/60i4yEH9vmZHiCs4VWNlxLA2/1S/SMmkWO/x6sDJ8MkdXFZ+ROzimlsEJhaNIgVcHFne47mOoUZvr1rxDjfSzjJ3+FNP/1ZNgXw+j04lyyD5fxPB3V/BY7tt0NW9mtmkIHxbcwYmVx9Ea2kQGEBnkbXRM4SCqMsdDY2NjdXx8fKVtT5Ri06foXx9nrz2KF33/zbvjh8j4qBN7eXYcQ/Y+SlvTIXJ7v4RvzwcBKLDaOJl5niBvd3mn5QKUUlu01rEX3y9j4NVNx7tQt/5AE/ezTD33KK9/PtvoROIKbdmwkjv3jaOlJQl149f/K28AD4uZ+sE+Ut4uTgq8OmrUD/O/luPl5cUbWY+Ttn6m0YnEZcpLmEvzJTfiZlLYxy6F5kONjiQMIAVeXYU1J+WmX0jQDam59H5Y8jTY5Go+VZ7dhl7xIt7zx7LXHsmZm37BI7Kt0amEQaTAq7EG9erxsPsLxAWMgA0fwszhkCPX1qyyctOwfn0das07fGvtzcouX9CyqawgWJ1JgVdjSim6NApnUs6t2IdNg+R4+OQqbIf/ICX7vNHxxIWSNlE07SrsR9fxpHU8eQPf4eFBcum86k4KvJrr3iiY9NxC9oQOgbt+A48aqK+H8fUb93P/N5s5cCbH6IjVm90Oa6agvxjEmXNFjDW9zNCxT/CvqxrIAlRCCry669YwGIA1iWchrAXZo5fxq+7Ko24/MObgA4x9dy6P/7gdm11Ov690Oadh1vWw4gWSw/sx+PxkHhp1E12jg41OJqoIKfBqLqyGJ43DfFmbeBaAWdsyuK9gAsk93ybW/TgrvJ4mf+scEpIy/vY17HbNz9tPcr7IVlmxXd+eBfBRFzi2Fq6ZwmNMpGbNYDrUCzQ6mahCpMAF3RuGsOlIOln5RXyx9ghXNQohove/UBPW4BbelKnuH+C38C7IPVvq8+P2p/Dg7G18ufZo5QZ3EXtOZtPzrThav7iMbi/8xC8vDYM5oyGwLty9mqQGN7PhSAYj20XIsIn4EylwQfdGNSmw2nlm3k5Scwq4p2fJov9BDbCMW8pM79HUP7sSPuwIu+bCRWfvzk8oXnNj9qbj2GWo5bLY7Jon5+4gt8DKU/UTWWJ5jAH2P5jpfiOFY5ZCSGPmlVwO77p2dQxOK6oaKXBBp/o1cTMrFu04Rcs6/nSNrvn/PzRbONXqPoYWTcZWIxJ+HAuzb4bM4qv7nCuwsnzPaSKDvDiensfqxNKP0kXpZq4/yunko/wc+ik3H34Kv+DabBswl39nD+fLDclorflpazJdGtQkIlDWNBF/JgUu8PGw0DaqeGx1Qq/ov7xN794omL22COK6fwsDXoEjf8CHnWDteyzbcZzzRXbeHNmamj7uzNpwzIhdMITWulxrq59Kz+bMsndY5fUYtc6shL7PwV1xdOjam37Nwnjvt4Ms2nGKY2l5jGwfUXHBhcuQAhcA3NIxkt5NQkq9uG37uoF4uZlZczgTuj4A922CBr1g+XN0WXotN/jvoXODIG6IjWTF3jOcysqv9PxGePC7BG6avuHKSvzwSuwf9+AJ9TVEdUbdux6umgTm4rVLnrumOVa7ZtKc7Xi7m2VNb1EqKXABwHVtI/hybEfMpr9+SOZhMdOxfhCrD5ZcCCIgEm6ZTeZ135BfaOOtgldQs65nTINzaOC7TUmVG94AW49nsHD7STYdSWfz0b+foXOhAquN9RvXcnDKYPh6GLogh19b/AevO+b+5WLDUTW9mdAzmkKbnUEx4fh4yMrP4q/kvwpRJlc1CuaVxXs5lZX/v+Vn5+bE8FrhG6zre4iQLe9SK7EfswL78NamEQxrU5stxzLYeSKLqxqF0K9ZqMvMoNBa89aS/QT7umO1az5fc5iO9YP+9vEp2edZsHIDIdve5xp7HHl4MrPGOHLbjGNcr2bwN7+XCb2iSckp4M5u9Ry0J8LZSYGLMuneqPjkkdUHz3JjbCQA8xNO0KROECH9h0G3O2Dte3TaMI051pX8+N6PTLMN5ZQpnK/XH6NXkxCev7YF9YN9DNyLirE2MY31h9N4/trmpOYUMG3VIY6n5RFV868fMq7bsp0j819ijCkOpUycajKKkCHPMso/9JLb8XQz89qIlo7YBeEiZAhFlEmTMD+CfT1Yc/AsRTY7MzccY0dyFsPblExt8w6C/i/CgwnsqT2CG93XsspzEvvazuPtnhbij2YwcMof/LQl2aE5cwscu6Ki1pq3lu6jToAXt3aKYnSXepiV4st1R/78wJR9MG8CnRb25iZzHAUtb8Pt4e1E3vo+nmUobyHKQgpclIlSiu4Na7JyfwoDpvzBv+fvon3dQK6/aHaE2b82re/+DPPDO1Gd78W8fxEjN97I1rpTGRe6n2fmbWf/acesrzJj3VHavrScpPQ8h7w+wNLdZ9ienMVD/RrhYTET7u/Jta1rM2dzEtn5BXBwOcy6AT7qhN4znzlqIK81/Ba/66eCv8zjFhVLClyUWe+moWSft+JhMfHZ6Fh+vKcLAd7upT/YLxwGvgoP74a+z+OecYgnMp7nN8tE1n/5BOfTKvaDzkOp55j8y14KbXYW7zz1p5+dL7LxYVwiWXlF5drGwTM5/HvBLqJDfBjR9v/L+O523txum49+v33x2iWntkOvpzh06waeyr+dJk1blGu7QvwdGQMXZTa0dW3qB/sQU9sfUymzVUrlHQRXPVI8/XDvz3iv/pQ7zszCPnU2NOwDMddD0yHgWeOKc1ltdibN2Y6nm5mIQC+W7Dr9/2eTAvO3neCtpfs5lHKOd25qc0Xb2HUii9FfbMJsUky7vT0Wax7sWQI7vqdp4gqecrOzo6ApMSM+x9R8KFjcWb22eFilS4Oal3h1Ia6MFLgoM6UUrSICruzJZjeIGUlgzEg+mrsc+9aZjDsZj1fiPWDxhOi+0GQwNB4Ivpc3RvzJH4dJSMrk/VvakpSex1tL9/9ptsyc+CRMCuZuO8HwtnXo0Tjksl5/6/EMxnyxiUiPfGb0yCFk5YTioRLrefCrDd0m8ptHX8YtzuJbn050tRS/K1l/KI3IIC+5KrxwmHIVuFJqEPAeYAY+01q/XiGphEv719C+XJfsyYzsUawc5YPP/vmwbzHsXwwoqN0W6vco/hPVGdz/OnNFa82xtDxWHUjl3RUHGNKyFte2qsXhs7m8tXQ/S3ed5o5u9UlMyWHr8UweHdCYudtO8Mz8nSyd2ANvdwtZeUXM25ZM26hAWkeW8g9TwTnOHdrAlh++5XvTDpoVJKKWa/ANh3ajocV1ENkJTGa6FtrwXbGCn7acoGt0MHa7ZuORdAa2CHP8L1RUW1dc4EopM/Ah0B9IBjYrpX7WWu+pqHDCNblbTEy+riXDP1rLf/bU5vlr34TBb8DpnXBgCRz6HdZ/AGvfRSsTmb4NOeLRjP2mBhw1RXGISPZlmUnOKD7js0mYHy8Pj0EpRXSIL41CfVmyu7jA58QnYzEpbu4YRYd6Qdw0fQPvLDtAdKgvby3dT3puIQC3t/HnkbaKoJz9xWPYJxPg9E58tY07taIgrC2q+VPQqB/UagumP3985OVuZkjLWizccZKXhrXgyNlcsvKL6BItwyfCccpzBN4RSNRaHwZQSn0HDAOkwMUltY4M4LZOUcxYd5Tr20fQorY/OrwlS9NC+OVkfw54niYscxvtTAdok3mI1qYVtFO5/3t+tjmQgrA6eAXXwyckCrV1M3gFgVcg99ZO46cdZ0nfU8DRLQncV9eb4FOrCS7I5p3ovexbv5A8lckHvrm0ispFpR/GZ18G7Ct5cc8AqNWK1Db38uhGbxrH9uWZEZ0uuU8j20fwfXwSS3adJiOv+B+GLg3k4gvCcdSVLsajlLoeGKS1/lfJ7VFAJ631/Rc9bjwwHiAqKqr9sWPVZ7Ej8c+y8oro8/ZKomp68+Gt7XhuwS5W7E0hrIYHbSIDaBURQPPaNWgY4ksdf09MOcnF86tT90JaImQmQVYSZJ+EosubOmg1e2H2r42qURuC6pPmWZcv9plYcCqI63p1ZmL/Jtz4yXqOns3l90m98Pd2u+Rraq3p+dZKIgK98HIzc/hsLnGP9rrC344Q/08ptUVrHXvx/eU5Ai9tGsJf/jXQWk8HpgPExsbKYtHif/y93Xj66mZM+mE7vd5aickEzw5pxthu9Utdk4WAqOI/jQf89WdF+ZCXDvkZaOt5Hpi5nsycXGp4efD+bbFYLO7g4Vc828XTH4u7759OYa8JTOxr5+y8XUyNO8Tv+1PZfTKbN0e2KlN5Q/GHvCPbRfDubwfwsJgY0U5WEBSOVZ4CTwYiL7gdAZwsXxxR3YxoV4fle86QX2TjpWEtqFvzCk+1d/MqPlHGvw4KqNXKm0WrjzChQzSWBk3L9hJmE6+PbEndYG/eXLKfNpEBfzlR6VJGtKvDlBUHOF9kl+mDwuHKU+CbgUZKqfrACeBm4NYKSSWqDaUUH49qX+Gve0NsJH8cOMutHaMuO8+9vRrSNTqYiECvss93LxEZ5E2n+kFsPJJOZylw4WBXXOBaa6tS6n5gKcXTCL/QWu+usGRClEPjMD+WPtzjip/fprRphWX0xOCmbDycToifxxW/hhBlUa554FrrX4BfKiiLEC6hXVQg7aLk6vHC8WQtFCGEcFJS4EII4aSkwIUQwklJgQshhJOSAhdCCCclBS6EEE5KClwIIZyUFLgQQjipK16N8Io2plQq4IzLEQYDZ40OUcmq4z5D9dzv6rjP4Fz7XVdr/ZdLSVVqgTsrpVR8aUs5urLquM9QPfe7Ou4zuMZ+yxCKEEI4KSlwIYRwUlLgZTPd6AAGqI77DNVzv6vjPoML7LeMgQshhJOSI3AhhHBSUuBCCOGkpMAvk1LqUaWUVkoFG53F0ZRSbyml9imldiil5imlrvwyNVWcUmqQUmq/UipRKfWk0Xkqg1IqUikVp5Taq5TarZR6yOhMlUUpZVZKbVNKLTI6S3lIgV8GpVQk0B84bnSWSrIciNFatwIOAE8ZnMchlFJm4ENgMNAcuEUp1dzYVJXCCkzSWjcDOgP3VZP9BngI2Gt0iPKSAr88U4DHgWrxya/WepnW2lpycwNweZdodx4dgUSt9WGtdSHwHTDM4EwOp7U+pbXeWvJ9DsWFVsfYVI6nlIoAhgCfGZ2lvKTAy0gpNRQ4obXebnQWg9wJ/Gp0CAepAyRdcDuZalBkF1JK1QPaAhuNTVIp3qX4QMxudJDyKtdFjV2NUmoFEF7Kj54BngYGVG4ix/unfdZaLyh5zDMUv92eVZnZKpEq5b5q8S4LQCnlC/wETNRaZxudx5GUUtcAKVrrLUqpXkbnKS8p8AtorfuVdr9SqiVQH9iulILioYStSqmOWuvTlRixwv3dPv+XUmoMcA3QV7vuSQPJQOQFtyOAkwZlqVRKKTeKy3uW1nqu0XkqQTdgqFLqasATqKGU+kZrfbvBua6InMhzBZRSR4FYrbWzrGR2RZRSg4B3gJ5a61Sj8ziKUspC8Ye0fYETwGbgVq31bkODOZgqPhqZAaRrrScanaeylRyBP6q1vsboLFdKxsDFP/kA8AOWK6USlFIfGx3IEUo+qL0fWErxB3lzXL28S3QDRgF9Sv5+E0qOTIWTkCNwIYRwUnIELoQQTkoKXAghnJQUuBBCOCkpcCGEcFJS4EII4aSkwIUQwklJgQshhJP6P+AnPAwdids2AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameter containing:\n",
      "tensor([[1.5040, 1.9692]], requires_grad=True)\n",
      "Parameter containing:\n",
      "tensor([0.1418], requires_grad=True)\n"
     ]
    }
   ],
   "source": [
    "train_and_test()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}