线性回归的简洁实现.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.0.0\n",
      "[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU'), PhysicalDevice(name='/physical_device:GPU:1', device_type='GPU')] [PhysicalDevice(name='/physical_device:CPU:0', device_type='CPU')]\n",
      "['/device:GPU:0', '/device:GPU:1']\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "\n",
    "print(tf.__version__)\n",
    "\n",
    "from tensorflow.python.client import device_lib\n",
    "import os\n",
    "os.environ['CUDA_VISIBLE_DEVICES'] = \"0, 1\"\n",
    "gpus = tf.config.experimental.list_physical_devices(device_type='GPU')\n",
    "cpus = tf.config.experimental.list_physical_devices(device_type='CPU')\n",
    "print(gpus, cpus)\n",
    "# 设置当前程序的可见设备范围\n",
    "tf.config.experimental.set_visible_devices(devices=gpus, device_type='GPU')\n",
    "\n",
    "# 设置仅在需要时申请：\n",
    "for gpu in gpus:\n",
    "    tf.config.experimental.set_memory_growth(gpu, True)\n",
    "    \n",
    "# 下面的方式是设置Tensorflow固定消耗GPU:0的2GB显存\n",
    "tf.config.experimental.set_virtual_device_configuration(\n",
    "    gpus[0],\n",
    "    [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=200)]\n",
    ")\n",
    "\n",
    "def get_available_gpus():\n",
    "  local_device_protos = device_lib.list_local_devices()\n",
    "  return [x.name for x in local_device_protos if x.device_type == 'GPU']\n",
    "\n",
    "print(get_available_gpus())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<tf.Tensor: id=710889, shape=(1, 2), dtype=float32, numpy=array([[ 2. , -3.4]], dtype=float32)>"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "with tf.device('/device:GPU:0'):\n",
    "    w = tf.constant([[2, -3.4]])\n",
    "    b = tf.constant([4.2])\n",
    "    x = tf.random.normal([1000, 2], mean=0, stddev=10)\n",
    "    e = tf.random.normal([1000, 2], mean=0, stddev=0.1)\n",
    "    W = tf.Variable(tf.constant([5, 1]))\n",
    "    B = tf.Variable(tf.constant([1]))\n",
    "w"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(<tf.Tensor: id=710896, shape=(1000, 2), dtype=float32, numpy=\n",
       " array([[11.904217 , 10.821163 ],\n",
       "        [ 3.4219682, -3.1656156],\n",
       "        [-4.010166 ,  1.2370993],\n",
       "        ...,\n",
       "        [-7.40069  ,  5.107273 ],\n",
       "        [ 4.918063 ,  1.1951212],\n",
       "        [ 6.251775 ,  9.301053 ]], dtype=float32)>,\n",
       " <tf.Tensor: id=710920, shape=(1000, 1), dtype=float32, numpy=\n",
       " array([[-8.78352070e+00],\n",
       "        [ 2.18070297e+01],\n",
       "        [-8.02647018e+00],\n",
       "        [-5.05666256e-01],\n",
       "        [-6.48605270e+01],\n",
       "        [ 9.50017929e+00],\n",
       "        [ 1.44045305e+01],\n",
       "        [ 2.50916481e+01],\n",
       "        [-1.00638485e+00],\n",
       "        [ 6.16355553e+01],\n",
       "        [-1.07517786e+01],\n",
       "        [ 6.02755737e+00],\n",
       "        [ 4.86500206e+01],\n",
       "        [-1.33696604e+01],\n",
       "        [ 1.03340054e+00],\n",
       "        [ 3.43312225e+01],\n",
       "        [ 2.38971443e+01],\n",
       "        [-3.98574257e+01],\n",
       "        [ 5.11106415e+01],\n",
       "        [-4.44288731e+00],\n",
       "        [-2.60181007e+01],\n",
       "        [-5.44732361e+01],\n",
       "        [-2.24977150e+01],\n",
       "        [-2.40252762e+01],\n",
       "        [-2.46596260e+01],\n",
       "        [ 1.65571976e+00],\n",
       "        [-9.00336838e+01],\n",
       "        [ 2.17636375e+01],\n",
       "        [-2.21117363e+01],\n",
       "        [ 5.07166939e+01],\n",
       "        [ 1.98917389e+01],\n",
       "        [ 3.60041924e+01],\n",
       "        [ 2.96862144e+01],\n",
       "        [-1.88308678e+01],\n",
       "        [ 4.61680336e+01],\n",
       "        [ 2.60550385e+01],\n",
       "        [-1.35035486e+01],\n",
       "        [-3.06102943e+01],\n",
       "        [ 3.47101097e+01],\n",
       "        [-4.48693428e+01],\n",
       "        [ 4.09630890e+01],\n",
       "        [-7.21935806e+01],\n",
       "        [-4.80305901e+01],\n",
       "        [-2.14732170e+01],\n",
       "        [-1.90561333e+01],\n",
       "        [ 6.91856670e+00],\n",
       "        [-5.86408901e+00],\n",
       "        [ 6.17875786e+01],\n",
       "        [ 2.22767944e+01],\n",
       "        [ 2.40737391e+00],\n",
       "        [ 1.54278374e+01],\n",
       "        [ 4.46854095e+01],\n",
       "        [ 2.29316235e+00],\n",
       "        [-2.46641998e+01],\n",
       "        [ 4.63711739e-01],\n",
       "        [-2.86674614e+01],\n",
       "        [-6.62199860e+01],\n",
       "        [-3.77366753e+01],\n",
       "        [ 5.66625023e+01],\n",
       "        [ 6.95191193e+01],\n",
       "        [ 2.70517044e+01],\n",
       "        [ 5.87358475e+00],\n",
       "        [-4.17465744e+01],\n",
       "        [ 1.05233173e+01],\n",
       "        [ 2.91885071e+01],\n",
       "        [ 4.85009422e+01],\n",
       "        [ 1.40338984e+01],\n",
       "        [-2.20435333e+01],\n",
       "        [ 4.11234570e+00],\n",
       "        [ 7.01585541e+01],\n",
       "        [-4.42057533e+01],\n",
       "        [-9.57449055e+00],\n",
       "        [ 2.57697792e+01],\n",
       "        [ 7.08591175e+00],\n",
       "        [-1.79232645e+00],\n",
       "        [-7.23592834e+01],\n",
       "        [-1.25892630e+01],\n",
       "        [-2.10863419e+01],\n",
       "        [ 3.75184975e+01],\n",
       "        [-5.20874500e+00],\n",
       "        [ 7.02495241e+00],\n",
       "        [-1.07976227e+01],\n",
       "        [-4.01645088e+01],\n",
       "        [-8.86263657e+01],\n",
       "        [ 4.53526974e-01],\n",
       "        [ 1.05025169e+02],\n",
       "        [-1.48062372e+01],\n",
       "        [ 1.83266373e+01],\n",
       "        [ 5.58239841e+00],\n",
       "        [ 1.98226814e+01],\n",
       "        [-5.22374153e+00],\n",
       "        [ 3.41874962e+01],\n",
       "        [ 4.06769524e+01],\n",
       "        [ 6.96728745e+01],\n",
       "        [ 9.73151169e+01],\n",
       "        [-1.39823713e+01],\n",
       "        [ 5.79044628e+00],\n",
       "        [-7.67846680e+01],\n",
       "        [ 6.38062172e+01],\n",
       "        [ 2.76289368e+01],\n",
       "        [ 4.19219551e+01],\n",
       "        [-7.37291336e+01],\n",
       "        [-1.93696823e+01],\n",
       "        [ 4.50880699e+01],\n",
       "        [-2.69240341e+01],\n",
       "        [ 6.20467834e+01],\n",
       "        [ 1.85678673e+01],\n",
       "        [ 8.55661583e+00],\n",
       "        [-5.53801727e+00],\n",
       "        [-3.46388168e+01],\n",
       "        [-1.00998449e+01],\n",
       "        [-4.06570473e+01],\n",
       "        [-4.74536972e+01],\n",
       "        [ 7.27493954e+00],\n",
       "        [-1.96699028e+01],\n",
       "        [ 5.62911987e+01],\n",
       "        [-2.68439751e+01],\n",
       "        [ 6.06760025e-01],\n",
       "        [ 1.30198822e+01],\n",
       "        [-6.91068573e+01],\n",
       "        [-5.39326096e+01],\n",
       "        [ 6.62803955e+01],\n",
       "        [-2.64051056e+01],\n",
       "        [-1.73738594e+01],\n",
       "        [ 7.56919718e+00],\n",
       "        [-1.41753893e+01],\n",
       "        [-8.17238312e+01],\n",
       "        [ 4.18306923e+01],\n",
       "        [ 5.82690239e+00],\n",
       "        [ 7.99846039e+01],\n",
       "        [-2.46103172e+01],\n",
       "        [-3.48991661e+01],\n",
       "        [-5.78573341e+01],\n",
       "        [ 6.20465393e+01],\n",
       "        [-5.58329582e+01],\n",
       "        [-4.61518173e+01],\n",
       "        [-8.22190857e+01],\n",
       "        [ 3.40306511e+01],\n",
       "        [ 4.47802467e+01],\n",
       "        [ 3.34295464e+00],\n",
       "        [ 3.23040543e+01],\n",
       "        [ 8.91214142e+01],\n",
       "        [-2.89573402e+01],\n",
       "        [ 1.87659950e+01],\n",
       "        [-2.04393845e+01],\n",
       "        [ 3.28867378e+01],\n",
       "        [ 5.74338951e+01],\n",
       "        [ 6.93831329e+01],\n",
       "        [-4.90110130e+01],\n",
       "        [-5.23873863e+01],\n",
       "        [-5.34232235e+00],\n",
       "        [-2.44474182e+01],\n",
       "        [ 5.65141640e+01],\n",
       "        [ 3.02860565e+01],\n",
       "        [ 1.05659800e+01],\n",
       "        [ 4.61135368e+01],\n",
       "        [ 2.14750504e+00],\n",
       "        [-1.19939709e+00],\n",
       "        [ 2.08645973e+01],\n",
       "        [ 4.13796234e+01],\n",
       "        [ 7.85442352e-01],\n",
       "        [ 1.95490189e+01],\n",
       "        [-2.48074989e+01],\n",
       "        [-1.59188395e+01],\n",
       "        [ 4.00826340e+01],\n",
       "        [ 2.20633125e+01],\n",
       "        [ 5.23300591e+01],\n",
       "        [ 3.45073662e+01],\n",
       "        [ 2.98643150e+01],\n",
       "        [ 1.11086483e+01],\n",
       "        [-7.76138401e+00],\n",
       "        [ 2.30820618e+01],\n",
       "        [-4.59555435e+01],\n",
       "        [ 1.44135408e+01],\n",
       "        [-8.26985550e+01],\n",
       "        [ 1.18792057e-01],\n",
       "        [ 4.38558884e+01],\n",
       "        [ 5.15472069e+01],\n",
       "        [ 2.04623146e+01],\n",
       "        [ 5.16368027e+01],\n",
       "        [-1.51691980e+01],\n",
       "        [ 2.94734230e+01],\n",
       "        [-3.26895790e+01],\n",
       "        [ 6.39072514e+00],\n",
       "        [ 4.06046448e+01],\n",
       "        [ 3.76602745e+01],\n",
       "        [-3.44979820e+01],\n",
       "        [-4.67441559e+01],\n",
       "        [ 4.48875999e+01],\n",
       "        [-2.09342422e+01],\n",
       "        [-1.46400881e+01],\n",
       "        [-5.07367439e+01],\n",
       "        [ 4.38780632e+01],\n",
       "        [-5.08388062e+01],\n",
       "        [ 3.39016914e+01],\n",
       "        [-6.50951538e+01],\n",
       "        [ 1.34814854e+01],\n",
       "        [ 2.16336365e+01],\n",
       "        [-3.42920265e+01],\n",
       "        [ 4.82272100e+00],\n",
       "        [-5.69875488e+01],\n",
       "        [-3.74294853e+01],\n",
       "        [-3.20869751e+01],\n",
       "        [ 8.00343513e+00],\n",
       "        [ 2.81314659e+01],\n",
       "        [ 9.00516224e+00],\n",
       "        [-6.58643188e+01],\n",
       "        [-5.73857765e+01],\n",
       "        [ 5.21897621e+01],\n",
       "        [-6.60876694e+01],\n",
       "        [ 1.99626884e+01],\n",
       "        [-2.55790710e+01],\n",
       "        [-2.45290642e+01],\n",
       "        [ 2.22436638e+01],\n",
       "        [ 5.44836388e+01],\n",
       "        [ 1.57977896e+01],\n",
       "        [-1.88298416e+01],\n",
       "        [-4.31849174e+01],\n",
       "        [ 2.45590019e+01],\n",
       "        [-3.00129852e+01],\n",
       "        [ 4.72272911e+01],\n",
       "        [-7.94486713e+00],\n",
       "        [ 3.43010178e+01],\n",
       "        [-8.24008274e+00],\n",
       "        [ 3.54744148e+01],\n",
       "        [-4.91117821e+01],\n",
       "        [ 1.77545128e+01],\n",
       "        [ 4.85555687e+01],\n",
       "        [-1.94610500e+00],\n",
       "        [-8.56516838e+00],\n",
       "        [-6.23965340e+01],\n",
       "        [ 4.52761555e+00],\n",
       "        [-8.58302307e+00],\n",
       "        [-2.80667353e+00],\n",
       "        [-2.11207695e+01],\n",
       "        [ 4.42884398e+00],\n",
       "        [ 5.01335564e+01],\n",
       "        [-2.68981667e+01],\n",
       "        [-3.50201797e+01],\n",
       "        [ 6.71057434e+01],\n",
       "        [ 6.84992371e+01],\n",
       "        [ 4.73417397e+01],\n",
       "        [-3.17939720e+01],\n",
       "        [ 2.36841011e+01],\n",
       "        [-8.87723236e+01],\n",
       "        [ 1.06073074e+01],\n",
       "        [-6.28029366e+01],\n",
       "        [ 4.93322678e+01],\n",
       "        [-1.52723055e+01],\n",
       "        [-2.10832291e+01],\n",
       "        [-8.25151215e+01],\n",
       "        [-3.05489845e+01],\n",
       "        [ 4.58714676e+01],\n",
       "        [ 4.61274910e+00],\n",
       "        [ 1.54570065e+01],\n",
       "        [ 5.54194527e+01],\n",
       "        [-7.83229733e+00],\n",
       "        [ 2.35189056e+01],\n",
       "        [ 2.28329315e+01],\n",
       "        [-2.64641647e+01],\n",
       "        [-1.25728025e+01],\n",
       "        [-2.19509811e+01],\n",
       "        [ 1.04025450e+01],\n",
       "        [ 5.05148773e+01],\n",
       "        [ 2.40805702e+01],\n",
       "        [ 1.30734015e+00],\n",
       "        [ 1.55635023e+01],\n",
       "        [-7.50684433e+01],\n",
       "        [ 5.88236198e+01],\n",
       "        [-2.63657188e+00],\n",
       "        [ 5.26511955e+01],\n",
       "        [-6.59380722e+01],\n",
       "        [-5.53641090e+01],\n",
       "        [ 4.08938313e+00],\n",
       "        [ 1.00756903e+01],\n",
       "        [-4.61071110e+00],\n",
       "        [ 3.32417908e+01],\n",
       "        [-5.73085442e+01],\n",
       "        [-5.48605461e+01],\n",
       "        [ 2.87573051e+01],\n",
       "        [ 1.85085182e+01],\n",
       "        [-3.42666626e+01],\n",
       "        [-3.73640823e+01],\n",
       "        [ 4.95707035e-01],\n",
       "        [-4.70180359e+01],\n",
       "        [-2.05815887e+01],\n",
       "        [-3.69428711e+01],\n",
       "        [-9.72196770e+00],\n",
       "        [ 2.20662842e+01],\n",
       "        [-1.39278002e+01],\n",
       "        [-1.17031355e+01],\n",
       "        [-3.32232552e+01],\n",
       "        [ 6.40675888e+01],\n",
       "        [-1.86086655e+01],\n",
       "        [ 2.42218666e+01],\n",
       "        [ 4.29218025e+01],\n",
       "        [ 4.20468102e+01],\n",
       "        [ 8.94623871e+01],\n",
       "        [ 4.37955284e+01],\n",
       "        [ 7.24910812e+01],\n",
       "        [-1.33103342e+01],\n",
       "        [-7.01592827e+00],\n",
       "        [-1.02263832e+01],\n",
       "        [-6.21679115e+01],\n",
       "        [ 3.74958382e+01],\n",
       "        [-3.81010475e+01],\n",
       "        [ 1.95088005e+01],\n",
       "        [ 3.75122414e+01],\n",
       "        [ 7.81997070e+01],\n",
       "        [-8.20792465e+01],\n",
       "        [ 3.45908661e+01],\n",
       "        [ 1.40703659e+01],\n",
       "        [ 9.68710709e+00],\n",
       "        [ 6.99064407e+01],\n",
       "        [ 4.33588295e+01],\n",
       "        [ 1.39591818e+01],\n",
       "        [-1.69343910e+01],\n",
       "        [ 1.14840288e+01],\n",
       "        [-3.71660080e+01],\n",
       "        [ 6.75258255e+01],\n",
       "        [ 1.62666283e+01],\n",
       "        [-4.85836487e+01],\n",
       "        [ 5.44191265e+00],\n",
       "        [ 9.06977367e+00],\n",
       "        [ 5.94023094e+01],\n",
       "        [ 6.75284481e+00],\n",
       "        [ 1.11759272e+01],\n",
       "        [ 4.70913658e+01],\n",
       "        [ 5.39782410e+01],\n",
       "        [ 5.44661407e+01],\n",
       "        [-3.97269630e+01],\n",
       "        [ 5.98354683e+01],\n",
       "        [-2.00613708e+01],\n",
       "        [ 4.74390984e-01],\n",
       "        [-2.20401001e+01],\n",
       "        [ 2.32028961e+01],\n",
       "        [ 2.69903946e+01],\n",
       "        [ 6.82759476e+00],\n",
       "        [-3.71236305e+01],\n",
       "        [-2.66763878e+01],\n",
       "        [ 2.01449661e+01],\n",
       "        [ 3.97802773e+01],\n",
       "        [ 5.75701370e+01],\n",
       "        [-5.64488029e+00],\n",
       "        [-8.02841282e+00],\n",
       "        [-5.28445549e+01],\n",
       "        [-4.53371773e+01],\n",
       "        [-6.25199165e+01],\n",
       "        [-4.74568443e+01],\n",
       "        [-2.10485878e+01],\n",
       "        [-4.48977623e+01],\n",
       "        [-2.82765541e+01],\n",
       "        [ 1.27530193e+01],\n",
       "        [ 2.03854790e+01],\n",
       "        [ 5.21045160e+00],\n",
       "        [ 6.94506531e+01],\n",
       "        [ 1.38278065e+01],\n",
       "        [ 1.16471224e+01],\n",
       "        [ 2.96871109e+01],\n",
       "        [-3.31294670e+01],\n",
       "        [ 1.27447824e+01],\n",
       "        [ 2.98502960e+01],\n",
       "        [ 4.49085922e+01],\n",
       "        [-8.38062057e+01],\n",
       "        [ 2.51722984e+01],\n",
       "        [ 4.86077957e+01],\n",
       "        [-1.81457138e+01],\n",
       "        [ 4.33860893e+01],\n",
       "        [-8.52567062e+01],\n",
       "        [ 5.18447971e+00],\n",
       "        [-3.72131882e+01],\n",
       "        [-2.94277611e+01],\n",
       "        [-2.72019691e+01],\n",
       "        [ 1.56120548e+01],\n",
       "        [-2.28757401e+01],\n",
       "        [ 3.33318748e+01],\n",
       "        [-4.04456425e+00],\n",
       "        [-3.40764122e+01],\n",
       "        [-1.57807779e+01],\n",
       "        [-2.19350204e+01],\n",
       "        [-1.79155922e+01],\n",
       "        [ 1.10100985e+01],\n",
       "        [-5.64109726e+01],\n",
       "        [ 1.55914783e+00],\n",
       "        [ 6.19308052e+01],\n",
       "        [ 5.84295654e+01],\n",
       "        [ 3.16839447e+01],\n",
       "        [ 1.66347084e+01],\n",
       "        [ 4.63205338e+01],\n",
       "        [-1.83170776e+01],\n",
       "        [ 2.10656395e+01],\n",
       "        [ 1.05195389e+01],\n",
       "        [ 8.81784668e+01],\n",
       "        [-3.19322319e+01],\n",
       "        [-4.69167519e+01],\n",
       "        [ 1.97108803e+01],\n",
       "        [ 2.78034668e+01],\n",
       "        [ 4.54725113e+01],\n",
       "        [-7.20067978e-01],\n",
       "        [ 3.28900146e+01],\n",
       "        [-5.12661648e+00],\n",
       "        [ 1.38648129e+00],\n",
       "        [ 4.74395370e+01],\n",
       "        [ 1.01562813e+02],\n",
       "        [ 2.46212482e+00],\n",
       "        [-4.30049858e+01],\n",
       "        [-4.70223961e+01],\n",
       "        [-5.19817924e+01],\n",
       "        [-2.46934395e+01],\n",
       "        [-1.48229551e+01],\n",
       "        [-2.09077148e+01],\n",
       "        [-1.24248686e+01],\n",
       "        [-3.11701965e+01],\n",
       "        [ 7.11385422e+01],\n",
       "        [-2.15130844e+01],\n",
       "        [-5.29436779e+00],\n",
       "        [ 2.70996208e+01],\n",
       "        [ 5.62459145e+01],\n",
       "        [ 1.29378576e+01],\n",
       "        [-1.00518255e+01],\n",
       "        [-9.18946266e+00],\n",
       "        [ 2.41018448e+01],\n",
       "        [-1.24249659e+01],\n",
       "        [-4.85210495e+01],\n",
       "        [-1.96612244e+01],\n",
       "        [-3.87426720e+01],\n",
       "        [ 6.01904602e+01],\n",
       "        [-1.66022072e+01],\n",
       "        [ 3.34128647e+01],\n",
       "        [ 5.55609703e+00],\n",
       "        [-4.02519894e+00],\n",
       "        [-5.16511879e+01],\n",
       "        [-6.50697098e+01],\n",
       "        [ 6.68732357e+00],\n",
       "        [ 1.72159147e+00],\n",
       "        [ 8.09649658e+00],\n",
       "        [ 4.39971542e+01],\n",
       "        [ 3.34202499e+01],\n",
       "        [-1.35340853e+01],\n",
       "        [-5.06014919e+00],\n",
       "        [-8.35918198e+01],\n",
       "        [ 3.64968777e+00],\n",
       "        [ 1.39686832e+01],\n",
       "        [-1.12267780e+01],\n",
       "        [ 4.63976364e+01],\n",
       "        [-5.13200493e+01],\n",
       "        [-4.20342636e+01],\n",
       "        [-4.87339554e+01],\n",
       "        [-7.43192139e+01],\n",
       "        [ 1.75111809e+01],\n",
       "        [ 3.57035470e+00],\n",
       "        [ 4.11378746e+01],\n",
       "        [-2.26459198e+01],\n",
       "        [ 1.26754522e+01],\n",
       "        [ 1.36975012e+01],\n",
       "        [ 2.12246132e+01],\n",
       "        [ 3.54055824e+01],\n",
       "        [ 2.23212700e+01],\n",
       "        [-4.20076084e+00],\n",
       "        [-3.18190231e+01],\n",
       "        [ 4.26783257e+01],\n",
       "        [-2.95253258e+01],\n",
       "        [ 9.17039013e+00],\n",
       "        [-5.30512333e+00],\n",
       "        [-3.31596708e+00],\n",
       "        [-7.65257359e+00],\n",
       "        [-4.50729103e+01],\n",
       "        [ 3.43885574e+01],\n",
       "        [ 3.88937950e-01],\n",
       "        [-1.26396456e+01],\n",
       "        [-3.57595482e+01],\n",
       "        [ 9.15372372e+00],\n",
       "        [-4.53612175e+01],\n",
       "        [-6.64804688e+01],\n",
       "        [ 3.38568840e+01],\n",
       "        [-2.23896217e+01],\n",
       "        [-3.61792183e+01],\n",
       "        [-7.04654121e+00],\n",
       "        [-1.27682695e+01],\n",
       "        [ 3.25724754e+01],\n",
       "        [-4.00627403e+01],\n",
       "        [ 3.21533737e+01],\n",
       "        [ 7.01106415e+01],\n",
       "        [-4.63327942e+01],\n",
       "        [-1.28307390e+01],\n",
       "        [ 5.14730034e+01],\n",
       "        [-5.18356400e+01],\n",
       "        [ 2.60344696e+01],\n",
       "        [-1.08159113e+01],\n",
       "        [-1.78734131e+01],\n",
       "        [-1.66868477e+01],\n",
       "        [-2.36922646e+01],\n",
       "        [-1.56679678e+01],\n",
       "        [-1.22258310e+01],\n",
       "        [-3.55091095e+01],\n",
       "        [-5.70624638e+00],\n",
       "        [ 1.44472532e+01],\n",
       "        [ 1.46511803e+01],\n",
       "        [-2.36512718e+01],\n",
       "        [ 2.10650921e-01],\n",
       "        [ 6.70449829e+00],\n",
       "        [ 4.23872089e+00],\n",
       "        [ 2.01588268e+01],\n",
       "        [ 7.94322252e+00],\n",
       "        [-1.87136154e+01],\n",
       "        [ 1.58724184e+01],\n",
       "        [ 3.04041290e+01],\n",
       "        [-4.74530945e+01],\n",
       "        [-1.17038507e+01],\n",
       "        [-3.17060318e+01],\n",
       "        [ 1.74195309e+01],\n",
       "        [-1.26113443e+01],\n",
       "        [ 5.15670433e+01],\n",
       "        [-6.42464638e+00],\n",
       "        [ 2.27329750e+01],\n",
       "        [-2.84606590e+01],\n",
       "        [-7.00638199e+00],\n",
       "        [ 2.88044090e+01],\n",
       "        [-2.97424889e+01],\n",
       "        [ 2.67104950e+01],\n",
       "        [ 7.54493332e+01],\n",
       "        [-2.04661465e+00],\n",
       "        [ 1.80886803e+01],\n",
       "        [-3.12047348e+01],\n",
       "        [-1.03687429e+01],\n",
       "        [ 3.59962120e+01],\n",
       "        [-4.34177856e+01],\n",
       "        [-3.02839355e+01],\n",
       "        [-4.76953964e+01],\n",
       "        [ 5.19912300e+01],\n",
       "        [ 3.04103508e+01],\n",
       "        [ 1.69069252e+01],\n",
       "        [-3.68536682e+01],\n",
       "        [ 1.84858208e+01],\n",
       "        [-2.82114906e+01],\n",
       "        [-2.20879631e+01],\n",
       "        [ 9.95772934e+00],\n",
       "        [-1.17361450e+00],\n",
       "        [-1.40314665e+01],\n",
       "        [ 7.93798523e+01],\n",
       "        [ 3.10212936e+01],\n",
       "        [-2.88201942e+01],\n",
       "        [-1.29395399e+01],\n",
       "        [ 6.28681850e+00],\n",
       "        [ 6.62827835e+01],\n",
       "        [ 1.00496769e+01],\n",
       "        [ 1.35054941e+01],\n",
       "        [ 3.44495277e+01],\n",
       "        [-6.67939758e+01],\n",
       "        [ 4.03481789e+01],\n",
       "        [-2.46652679e+01],\n",
       "        [-1.37188997e+01],\n",
       "        [-3.88294678e+01],\n",
       "        [ 2.56613922e+01],\n",
       "        [ 4.55733910e+01],\n",
       "        [-3.81423492e+01],\n",
       "        [-5.72739754e+01],\n",
       "        [-1.39125757e+01],\n",
       "        [ 6.31996765e+01],\n",
       "        [ 3.77970695e-01],\n",
       "        [ 1.25250320e+01],\n",
       "        [-4.66940613e+01],\n",
       "        [ 2.35971756e+01],\n",
       "        [-6.87852783e+01],\n",
       "        [ 1.35864344e+01],\n",
       "        [-1.44256430e+01],\n",
       "        [-6.03880882e+01],\n",
       "        [-2.68375511e+01],\n",
       "        [ 4.03476753e+01],\n",
       "        [ 8.45277119e+00],\n",
       "        [ 4.67082176e+01],\n",
       "        [ 7.54009171e+01],\n",
       "        [ 3.36197586e+01],\n",
       "        [-2.33386116e+01],\n",
       "        [ 2.87871323e+01],\n",
       "        [ 1.66685390e+00],\n",
       "        [ 6.10147896e+01],\n",
       "        [-3.14346848e+01],\n",
       "        [-6.64979553e+01],\n",
       "        [ 7.26486969e+01],\n",
       "        [-6.52503204e+00],\n",
       "        [ 1.06074834e+00],\n",
       "        [-4.67649651e+01],\n",
       "        [-4.85610542e+01],\n",
       "        [ 2.50402527e+01],\n",
       "        [-5.63813286e+01],\n",
       "        [ 2.18378735e+00],\n",
       "        [ 1.12413998e+01],\n",
       "        [-3.93948631e+01],\n",
       "        [ 5.56325455e+01],\n",
       "        [-6.72707520e+01],\n",
       "        [ 1.73651009e+01],\n",
       "        [ 2.78529263e+00],\n",
       "        [-3.18312798e+01],\n",
       "        [ 6.02728500e+01],\n",
       "        [-4.83524437e+01],\n",
       "        [ 3.71740646e+01],\n",
       "        [-1.10113726e+01],\n",
       "        [-3.78842964e+01],\n",
       "        [ 5.83821449e+01],\n",
       "        [ 6.28348351e-01],\n",
       "        [ 4.50594368e+01],\n",
       "        [ 2.11367226e+01],\n",
       "        [-2.67294235e+01],\n",
       "        [-4.89256134e+01],\n",
       "        [-6.46220474e+01],\n",
       "        [-3.82224922e+01],\n",
       "        [-2.93778038e+01],\n",
       "        [ 4.47211189e+01],\n",
       "        [ 2.69442511e+00],\n",
       "        [ 6.52684631e+01],\n",
       "        [-8.76376152e-01],\n",
       "        [-4.53177986e+01],\n",
       "        [-2.47503281e+01],\n",
       "        [-6.52200165e+01],\n",
       "        [ 4.99431610e-01],\n",
       "        [ 3.74707909e+01],\n",
       "        [-2.13379211e+01],\n",
       "        [-3.24127197e+01],\n",
       "        [ 9.91700897e+01],\n",
       "        [-6.03997688e+01],\n",
       "        [ 5.59980431e+01],\n",
       "        [ 9.32062302e+01],\n",
       "        [ 1.65251884e+01],\n",
       "        [ 5.67505360e+00],\n",
       "        [ 3.35602036e+01],\n",
       "        [-4.36575031e+00],\n",
       "        [ 1.90106850e+01],\n",
       "        [ 1.13775892e+01],\n",
       "        [-3.83360062e+01],\n",
       "        [-1.13463650e+01],\n",
       "        [-1.04626854e+02],\n",
       "        [-6.22324638e+01],\n",
       "        [ 7.48346558e+01],\n",
       "        [ 3.75386429e+01],\n",
       "        [ 1.42398977e+01],\n",
       "        [-9.35002518e+00],\n",
       "        [-7.22932510e+01],\n",
       "        [-1.16366320e+01],\n",
       "        [-5.89081192e+01],\n",
       "        [ 3.23011093e+01],\n",
       "        [ 1.13343229e+01],\n",
       "        [-7.67106056e-01],\n",
       "        [ 1.23900576e+01],\n",
       "        [ 3.30350761e+01],\n",
       "        [ 1.65847473e+01],\n",
       "        [-3.14831314e+01],\n",
       "        [ 8.12647858e+01],\n",
       "        [ 2.87546730e+01],\n",
       "        [-1.46042728e+00],\n",
       "        [-5.05028572e+01],\n",
       "        [-3.38545227e+01],\n",
       "        [-3.10608063e+01],\n",
       "        [-1.52704344e+01],\n",
       "        [-9.72496414e+00],\n",
       "        [ 5.63076897e+01],\n",
       "        [-6.73669586e+01],\n",
       "        [ 2.44982719e-01],\n",
       "        [ 1.51601114e+01],\n",
       "        [-6.10637856e+00],\n",
       "        [-4.15165806e+00],\n",
       "        [ 5.23849678e+01],\n",
       "        [-7.23364830e-01],\n",
       "        [ 2.29512062e+01],\n",
       "        [ 2.14426804e+01],\n",
       "        [-1.57659502e+01],\n",
       "        [ 1.08287544e+01],\n",
       "        [-1.82203598e+01],\n",
       "        [ 3.19774857e+01],\n",
       "        [-8.80235901e+01],\n",
       "        [-6.05059052e+00],\n",
       "        [-6.24672031e+00],\n",
       "        [ 5.07657280e+01],\n",
       "        [ 3.53441620e+00],\n",
       "        [ 1.79414101e+01],\n",
       "        [-3.32435265e+01],\n",
       "        [ 1.95276356e+01],\n",
       "        [-8.29893036e+01],\n",
       "        [-3.37561798e+01],\n",
       "        [ 1.96379623e+01],\n",
       "        [ 4.08535690e+01],\n",
       "        [-7.41316319e+00],\n",
       "        [-1.49126081e+01],\n",
       "        [ 6.68853283e+00],\n",
       "        [-4.91862011e+00],\n",
       "        [ 4.74047432e+01],\n",
       "        [-8.48802261e+01],\n",
       "        [-1.55378437e+00],\n",
       "        [ 8.03047466e+00],\n",
       "        [-9.18181133e+00],\n",
       "        [-5.84433651e+00],\n",
       "        [-2.66408691e+01],\n",
       "        [ 2.82193565e+01],\n",
       "        [ 9.05837059e-01],\n",
       "        [ 1.49596510e+01],\n",
       "        [ 7.63746338e+01],\n",
       "        [ 4.91214333e+01],\n",
       "        [-2.26692924e+01],\n",
       "        [ 3.45801201e+01],\n",
       "        [-1.23401117e+01],\n",
       "        [ 4.90098686e+01],\n",
       "        [ 1.39958620e+01],\n",
       "        [ 4.89827232e+01],\n",
       "        [ 1.05435905e+02],\n",
       "        [ 1.81063499e+01],\n",
       "        [ 7.05704975e+00],\n",
       "        [-4.78464394e+01],\n",
       "        [ 4.17720375e+01],\n",
       "        [ 4.25575829e+01],\n",
       "        [-6.23054790e+00],\n",
       "        [ 2.98195801e+01],\n",
       "        [ 7.08381805e+01],\n",
       "        [ 5.47110558e-01],\n",
       "        [-4.46177864e+01],\n",
       "        [ 7.62147675e+01],\n",
       "        [-2.61414642e+01],\n",
       "        [ 1.02692127e+01],\n",
       "        [ 4.94924355e+01],\n",
       "        [ 4.68931770e+01],\n",
       "        [ 9.33044739e+01],\n",
       "        [ 5.60724373e+01],\n",
       "        [-2.74400291e+01],\n",
       "        [ 6.95132980e+01],\n",
       "        [ 2.38558540e+01],\n",
       "        [-9.86353874e+00],\n",
       "        [-4.56137276e+01],\n",
       "        [-6.49061356e+01],\n",
       "        [ 6.97650051e+00],\n",
       "        [ 1.53028822e+01],\n",
       "        [-2.66923561e+01],\n",
       "        [-3.09972343e+01],\n",
       "        [ 4.38746872e+01],\n",
       "        [-3.68726959e+01],\n",
       "        [ 1.13230381e+01],\n",
       "        [-8.00331573e+01],\n",
       "        [-3.01683960e+01],\n",
       "        [-9.34112453e+00],\n",
       "        [-1.51566286e+01],\n",
       "        [-3.75595093e+00],\n",
       "        [ 5.38045158e+01],\n",
       "        [-1.35741148e+01],\n",
       "        [-2.17240944e+01],\n",
       "        [ 4.87050171e+01],\n",
       "        [ 1.28507528e+01],\n",
       "        [ 2.59341736e+01],\n",
       "        [-2.42731705e+01],\n",
       "        [ 8.51748562e+00],\n",
       "        [ 2.58961868e+01],\n",
       "        [ 2.11765709e+01],\n",
       "        [ 1.65423565e+01],\n",
       "        [-2.07333107e+01],\n",
       "        [-8.18617630e+01],\n",
       "        [-9.11011982e+00],\n",
       "        [ 3.34849091e+01],\n",
       "        [-1.11036119e+01],\n",
       "        [-1.12950325e-01],\n",
       "        [ 3.49630623e+01],\n",
       "        [ 4.83903122e+01],\n",
       "        [-4.11932945e+01],\n",
       "        [ 3.83818970e+01],\n",
       "        [-5.39906082e+01],\n",
       "        [-3.67931404e+01],\n",
       "        [-6.44142914e+01],\n",
       "        [ 4.37484856e+01],\n",
       "        [-2.21891518e+01],\n",
       "        [ 5.52602806e+01],\n",
       "        [ 2.06695328e+01],\n",
       "        [-1.12412964e+02],\n",
       "        [ 4.81384277e+01],\n",
       "        [ 7.01509476e-01],\n",
       "        [ 7.26484585e+00],\n",
       "        [-2.93831940e+01],\n",
       "        [-1.34028597e+01],\n",
       "        [ 1.12223907e+02],\n",
       "        [-6.01141453e+00],\n",
       "        [ 2.34520226e+01],\n",
       "        [ 5.47411442e-01],\n",
       "        [ 4.59358559e+01],\n",
       "        [-4.14912987e+01],\n",
       "        [-1.04770298e+01],\n",
       "        [ 2.94863968e+01],\n",
       "        [ 8.21494961e+00],\n",
       "        [-5.60502720e+00],\n",
       "        [-4.95457935e+00],\n",
       "        [-3.83527489e+01],\n",
       "        [ 7.48631439e+01],\n",
       "        [ 1.95046043e+01],\n",
       "        [-5.20973282e+01],\n",
       "        [ 2.46970291e+01],\n",
       "        [ 5.55860710e+01],\n",
       "        [ 1.37504930e+01],\n",
       "        [ 8.10544014e-01],\n",
       "        [ 7.01055374e+01],\n",
       "        [-8.73673248e+01],\n",
       "        [-1.96002769e+00],\n",
       "        [-9.82352924e+00],\n",
       "        [ 7.30241013e+00],\n",
       "        [ 6.26011009e+01],\n",
       "        [-7.12784290e+00],\n",
       "        [ 2.67502060e+01],\n",
       "        [-6.88575745e+01],\n",
       "        [ 1.87421799e+00],\n",
       "        [-4.24094810e+01],\n",
       "        [ 3.92963142e+01],\n",
       "        [ 8.81023102e+01],\n",
       "        [ 8.65426025e+01],\n",
       "        [ 7.82828827e+01],\n",
       "        [-1.74539490e+01],\n",
       "        [-6.44924774e+01],\n",
       "        [ 6.15357494e+00],\n",
       "        [ 3.99560928e-02],\n",
       "        [-1.20517817e+01],\n",
       "        [-9.68422985e+00],\n",
       "        [ 5.23512115e+01],\n",
       "        [-4.94601631e+01],\n",
       "        [-1.19188623e+01],\n",
       "        [-1.45807848e+01],\n",
       "        [ 2.84018135e+01],\n",
       "        [ 6.78411331e+01],\n",
       "        [-2.00633583e+01],\n",
       "        [ 5.87710381e+01],\n",
       "        [-1.46930361e+01],\n",
       "        [-5.16858139e+01],\n",
       "        [-7.57704449e+00],\n",
       "        [ 6.57344723e+00],\n",
       "        [-9.06222630e+00],\n",
       "        [-1.90497208e+01],\n",
       "        [ 4.01422081e+01],\n",
       "        [-5.88963509e+00],\n",
       "        [ 1.40990686e+01],\n",
       "        [-2.42439079e+01],\n",
       "        [ 4.13441696e+01],\n",
       "        [ 5.71745539e+00],\n",
       "        [-8.91312408e+00],\n",
       "        [ 4.91745949e+01],\n",
       "        [ 6.52304993e+01],\n",
       "        [ 1.06008787e+01],\n",
       "        [-1.80862732e+01],\n",
       "        [-4.08831501e+00],\n",
       "        [-1.19479723e+01],\n",
       "        [ 3.16782188e+00],\n",
       "        [-1.15011692e+01],\n",
       "        [ 3.62170334e+01],\n",
       "        [-1.39846411e+01],\n",
       "        [ 5.59147682e+01],\n",
       "        [-3.68395729e+01],\n",
       "        [-1.73639603e+01],\n",
       "        [ 6.92887878e+01],\n",
       "        [-2.93351021e+01],\n",
       "        [-2.20175591e+01],\n",
       "        [ 6.03887482e+01],\n",
       "        [-1.13360023e+01],\n",
       "        [-1.87334404e+01],\n",
       "        [-4.55627556e+01],\n",
       "        [ 3.42326164e+01],\n",
       "        [ 4.51385040e+01],\n",
       "        [ 7.67571487e+01],\n",
       "        [-2.25933113e+01],\n",
       "        [-4.38360882e+00],\n",
       "        [ 5.51101780e+00],\n",
       "        [ 5.02591057e+01],\n",
       "        [-3.44982300e+01],\n",
       "        [-5.94454479e+00],\n",
       "        [-2.41742859e+01],\n",
       "        [-1.02258263e+02],\n",
       "        [ 4.43758736e+01],\n",
       "        [ 1.14544601e+01],\n",
       "        [-1.17045853e+02],\n",
       "        [ 9.90640640e+01],\n",
       "        [ 1.66997280e+01],\n",
       "        [ 2.97133331e+01],\n",
       "        [-4.96173439e+01],\n",
       "        [-1.09206238e+02],\n",
       "        [ 9.97029572e+01],\n",
       "        [-6.61259174e+00],\n",
       "        [-7.45405350e+01],\n",
       "        [-3.97598915e+01],\n",
       "        [ 2.59244194e+01],\n",
       "        [ 4.92736015e+01],\n",
       "        [-2.29078369e+01],\n",
       "        [ 4.48144722e+01],\n",
       "        [ 2.55942383e+01],\n",
       "        [ 1.43132210e-01],\n",
       "        [ 7.69331551e+00],\n",
       "        [ 1.50475855e+01],\n",
       "        [ 2.18172684e+01],\n",
       "        [ 4.03515816e-01],\n",
       "        [-7.23280907e+00],\n",
       "        [ 1.76002731e+01],\n",
       "        [ 7.19112396e+01],\n",
       "        [-1.30886841e+00],\n",
       "        [ 2.12647705e+01],\n",
       "        [ 1.96075439e-03],\n",
       "        [ 6.05255470e+01],\n",
       "        [-1.34194956e+01],\n",
       "        [ 4.60443459e+01],\n",
       "        [-1.17948599e+01],\n",
       "        [-7.67851925e+00],\n",
       "        [ 6.25464630e+00],\n",
       "        [-4.23093319e+00],\n",
       "        [ 2.35304680e+01],\n",
       "        [ 4.11152000e+01],\n",
       "        [-6.81381836e+01],\n",
       "        [ 9.14747620e+01],\n",
       "        [-8.55778275e+01],\n",
       "        [ 3.27091751e+01],\n",
       "        [-6.64377213e-01],\n",
       "        [-7.17252655e+01],\n",
       "        [ 5.61712608e+01],\n",
       "        [-1.32365980e+01],\n",
       "        [ 2.33207359e+01],\n",
       "        [ 3.06992645e+01],\n",
       "        [ 1.92503033e+01],\n",
       "        [ 2.47797394e+00],\n",
       "        [ 9.87818432e+00],\n",
       "        [ 2.29573059e+01],\n",
       "        [ 7.06560612e+00],\n",
       "        [ 4.01603699e+00],\n",
       "        [-5.47295666e+00],\n",
       "        [-4.14559898e+01],\n",
       "        [-9.51694393e+00],\n",
       "        [ 3.34808769e+01],\n",
       "        [-1.94933434e+01],\n",
       "        [-1.97045097e+01],\n",
       "        [ 1.77536926e+01],\n",
       "        [-4.09665394e+00],\n",
       "        [ 1.52266779e+01],\n",
       "        [-2.09734001e+01],\n",
       "        [-4.14247627e+01],\n",
       "        [-2.51859398e+01],\n",
       "        [ 7.84740067e+00],\n",
       "        [-4.96022377e+01],\n",
       "        [-4.49795723e+01],\n",
       "        [ 4.90773506e+01],\n",
       "        [-1.01697330e+01],\n",
       "        [-5.71050873e+01],\n",
       "        [ 4.49118805e+00],\n",
       "        [-6.13253880e+00],\n",
       "        [ 4.15971756e+01],\n",
       "        [ 2.92127991e+00],\n",
       "        [ 6.00408783e+01],\n",
       "        [-1.43964186e+01],\n",
       "        [ 1.32020330e+01],\n",
       "        [-4.29118500e+01],\n",
       "        [ 1.21284266e+01],\n",
       "        [-4.75254517e+01],\n",
       "        [-6.90408802e+00],\n",
       "        [-4.32060051e+01],\n",
       "        [ 1.35120211e+01],\n",
       "        [ 1.34445333e+01],\n",
       "        [-4.64105988e+01],\n",
       "        [-1.12136459e+00],\n",
       "        [ 3.56990623e+01],\n",
       "        [-6.80308990e+01],\n",
       "        [-5.59935284e+00],\n",
       "        [-7.27918339e+00],\n",
       "        [ 6.91030645e+00],\n",
       "        [ 3.25201988e+01],\n",
       "        [-8.25775757e+01],\n",
       "        [ 1.25050669e+01],\n",
       "        [-6.91451359e+00],\n",
       "        [-8.61675167e+00],\n",
       "        [-8.18287506e+01],\n",
       "        [ 2.85496292e+01],\n",
       "        [-2.29738388e+01],\n",
       "        [-2.91973534e+01],\n",
       "        [ 5.39379234e+01],\n",
       "        [-2.88215923e+00],\n",
       "        [ 3.05855331e+01],\n",
       "        [ 4.48371658e+01],\n",
       "        [-4.65450478e+01],\n",
       "        [-4.59851952e+01],\n",
       "        [ 8.04877090e+00],\n",
       "        [-5.67130737e+01],\n",
       "        [ 2.69440079e+01],\n",
       "        [-1.36632013e+01],\n",
       "        [-3.93351707e+01],\n",
       "        [-2.56468487e+00],\n",
       "        [-2.36065292e+01],\n",
       "        [ 6.52849960e+01],\n",
       "        [-1.98504353e+00],\n",
       "        [ 4.29475365e+01],\n",
       "        [ 3.28429565e+01],\n",
       "        [ 4.08384628e+01],\n",
       "        [-2.85039997e+00],\n",
       "        [ 1.69373856e+01],\n",
       "        [-5.77991676e+01],\n",
       "        [ 9.24762344e+01],\n",
       "        [ 7.39179153e+01],\n",
       "        [-3.63774490e+01],\n",
       "        [ 1.16271286e+01],\n",
       "        [-1.63470688e+01],\n",
       "        [ 2.33097496e+01],\n",
       "        [ 4.74442825e+01],\n",
       "        [-3.10414619e+01],\n",
       "        [-3.15517464e+01],\n",
       "        [ 4.70622749e+01],\n",
       "        [-2.79661064e+01],\n",
       "        [ 9.97271347e+00],\n",
       "        [-1.49200315e+01]], dtype=float32)>)"
      ]
     },
     "execution_count": 93,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import random\n",
    "from matplotlib import pyplot as plt\n",
    "# 线性回归模型， y = \n",
    "# 生成数据,生成1000组数据\n",
    "\n",
    "num_inputs = 2\n",
    "num_examples = 1000\n",
    "true_y = tf.matmul(x, tf.transpose(w)) + b\n",
    "x, true_y\n",
    "# 读取数据"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7ff59c368b38>"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 252x180 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def set_figsize(figsize=(3.5, 2.5)):\n",
    "    plt.rcParams['figure.figsize'] = figsize\n",
    "\n",
    "set_figsize()\n",
    "\n",
    "plt.scatter(x[: ,1], true_y, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [],
   "source": [
    "def data_scale(x, y):\n",
    "    return x, y\n",
    "\n",
    "db = tf.data.Dataset.from_tensor_slices((x, true_y))\n",
    "db_all = db.map(data_scale)\n",
    "# 为了每个epchoes都打散，可以把下面　db　放到每轮内部去shuffle和ｂａｔｃｈ\n",
    "# db = db_all.shuffle(10)\n",
    "# db_batch = db.batch(32)\n",
    "# 查看一组数据\n",
    "# print(next(iter(db_batch)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 构建模型（此处就不需要自己定义变量在进行矩阵计算了，直接调用包就好了）\n",
    "# 导入模块\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "from tensorflow import initializers as init\n",
    "\n",
    "# 使用Sequential模式创建\n",
    "model = keras.Sequential()\n",
    "model.add(layers.Dense(1, kernel_initializer=init.RandomNormal(stddev=0.01)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 定义损失函数\n",
    "from tensorflow import losses\n",
    "\n",
    "#均方损失 \n",
    "loss = losses.MeanSquaredError()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 定义优化算法\n",
    "from tensorflow.keras import optimizers\n",
    "trainer = optimizers.SGD(learning_rate=0.001)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "在使用Tensorflow训练模型时，我们通过调用tensorflow.GradientTape记录动态图梯度，执行tape.gradient获得动态图中各变量梯度。通过 model.trainable_variables 找到需要更新的变量，并用 trainer.apply_gradients 更新权重，完成一步训练。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x.shape (1000, 2)\n",
      "0 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 0.17276232],\n",
      "       [-0.46700823]], dtype=float32)>\n",
      "[array([[ 0.17276232],\n",
      "       [-0.46700823]], dtype=float32), array([0.0116694], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "1 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 0.6112487],\n",
      "       [-0.8874874]], dtype=float32)>\n",
      "[array([[ 0.6112487],\n",
      "       [-0.8874874]], dtype=float32), array([-0.0001864], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "2 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.0133995],\n",
      "       [-1.6656827]], dtype=float32)>\n",
      "[array([[ 1.0133995],\n",
      "       [-1.6656827]], dtype=float32), array([0.0126952], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "3 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.056464 ],\n",
      "       [-2.1160405]], dtype=float32)>\n",
      "[array([[ 1.056464 ],\n",
      "       [-2.1160405]], dtype=float32), array([0.01388448], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "4 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.3288705],\n",
      "       [-2.3778358]], dtype=float32)>\n",
      "[array([[ 1.3288705],\n",
      "       [-2.3778358]], dtype=float32), array([0.02582193], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "5 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.4296857],\n",
      "       [-2.5542493]], dtype=float32)>\n",
      "[array([[ 1.4296857],\n",
      "       [-2.5542493]], dtype=float32), array([0.0382039], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "6 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.5422536],\n",
      "       [-2.711399 ]], dtype=float32)>\n",
      "[array([[ 1.5422536],\n",
      "       [-2.711399 ]], dtype=float32), array([0.03969093], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "7 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.594546 ],\n",
      "       [-2.8425722]], dtype=float32)>\n",
      "[array([[ 1.594546 ],\n",
      "       [-2.8425722]], dtype=float32), array([0.04418265], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "8 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.6348226],\n",
      "       [-2.9606261]], dtype=float32)>\n",
      "[array([[ 1.6348226],\n",
      "       [-2.9606261]], dtype=float32), array([0.05111372], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "9 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.7649112],\n",
      "       [-3.0566697]], dtype=float32)>\n",
      "[array([[ 1.7649112],\n",
      "       [-3.0566697]], dtype=float32), array([0.06130499], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "10 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.8110687],\n",
      "       [-3.1093307]], dtype=float32)>\n",
      "[array([[ 1.8110687],\n",
      "       [-3.1093307]], dtype=float32), array([0.0698735], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "11 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.8230801],\n",
      "       [-3.167548 ]], dtype=float32)>\n",
      "[array([[ 1.8230801],\n",
      "       [-3.167548 ]], dtype=float32), array([0.07681113], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "12 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.8610773],\n",
      "       [-3.2136152]], dtype=float32)>\n",
      "[array([[ 1.8610773],\n",
      "       [-3.2136152]], dtype=float32), array([0.0853779], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "13 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.8733169],\n",
      "       [-3.2371953]], dtype=float32)>\n",
      "[array([[ 1.8733169],\n",
      "       [-3.2371953]], dtype=float32), array([0.09254031], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "14 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.8724623],\n",
      "       [-3.2463076]], dtype=float32)>\n",
      "[array([[ 1.8724623],\n",
      "       [-3.2463076]], dtype=float32), array([0.09995545], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "15 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.8668494],\n",
      "       [-3.2509267]], dtype=float32)>\n",
      "[array([[ 1.8668494],\n",
      "       [-3.2509267]], dtype=float32), array([0.10725895], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "16 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.924314 ],\n",
      "       [-3.2592325]], dtype=float32)>\n",
      "[array([[ 1.924314 ],\n",
      "       [-3.2592325]], dtype=float32), array([0.11595054], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "17 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.924371 ],\n",
      "       [-3.2812204]], dtype=float32)>\n",
      "[array([[ 1.924371 ],\n",
      "       [-3.2812204]], dtype=float32), array([0.12355934], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "18 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9179723],\n",
      "       [-3.3031604]], dtype=float32)>\n",
      "[array([[ 1.9179723],\n",
      "       [-3.3031604]], dtype=float32), array([0.13105816], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "19 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9218725],\n",
      "       [-3.3201401]], dtype=float32)>\n",
      "[array([[ 1.9218725],\n",
      "       [-3.3201401]], dtype=float32), array([0.13899106], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "20 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9454665],\n",
      "       [-3.332394 ]], dtype=float32)>\n",
      "[array([[ 1.9454665],\n",
      "       [-3.332394 ]], dtype=float32), array([0.14701949], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "21 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9716191],\n",
      "       [-3.3371816]], dtype=float32)>\n",
      "[array([[ 1.9716191],\n",
      "       [-3.3371816]], dtype=float32), array([0.15527123], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "22 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9921606],\n",
      "       [-3.3592355]], dtype=float32)>\n",
      "[array([[ 1.9921606],\n",
      "       [-3.3592355]], dtype=float32), array([0.16356605], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "23 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0071225],\n",
      "       [-3.3498116]], dtype=float32)>\n",
      "[array([[ 2.0071225],\n",
      "       [-3.3498116]], dtype=float32), array([0.17146839], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "24 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.988313 ],\n",
      "       [-3.3771398]], dtype=float32)>\n",
      "[array([[ 1.988313 ],\n",
      "       [-3.3771398]], dtype=float32), array([0.17980638], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "25 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9880025],\n",
      "       [-3.386332 ]], dtype=float32)>\n",
      "[array([[ 1.9880025],\n",
      "       [-3.386332 ]], dtype=float32), array([0.18786377], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "26 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9932159],\n",
      "       [-3.388912 ]], dtype=float32)>\n",
      "[array([[ 1.9932159],\n",
      "       [-3.388912 ]], dtype=float32), array([0.19589522], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "27 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9889312],\n",
      "       [-3.384678 ]], dtype=float32)>\n",
      "[array([[ 1.9889312],\n",
      "       [-3.384678 ]], dtype=float32), array([0.20387168], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "28 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0084288],\n",
      "       [-3.3831859]], dtype=float32)>\n",
      "[array([[ 2.0084288],\n",
      "       [-3.3831859]], dtype=float32), array([0.21189724], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "29 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9827847],\n",
      "       [-3.3643208]], dtype=float32)>\n",
      "[array([[ 1.9827847],\n",
      "       [-3.3643208]], dtype=float32), array([0.21982917], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "30 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9956055],\n",
      "       [-3.3616626]], dtype=float32)>\n",
      "[array([[ 1.9956055],\n",
      "       [-3.3616626]], dtype=float32), array([0.2277532], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "31 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0577025],\n",
      "       [-3.3619459]], dtype=float32)>\n",
      "[array([[ 2.0577025],\n",
      "       [-3.3619459]], dtype=float32), array([0.23571971], dtype=float32)]\n",
      "\n",
      " epoch 0 , loss: 16.048676\n",
      "x.shape (1000, 2)\n",
      "0 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0641348],\n",
      "       [-3.3466244]], dtype=float32)>\n",
      "[array([[ 2.0641348],\n",
      "       [-3.3466244]], dtype=float32), array([0.24328531], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "1 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0473402],\n",
      "       [-3.3492382]], dtype=float32)>\n",
      "[array([[ 2.0473402],\n",
      "       [-3.3492382]], dtype=float32), array([0.25113416], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "2 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0359929],\n",
      "       [-3.3847485]], dtype=float32)>\n",
      "[array([[ 2.0359929],\n",
      "       [-3.3847485]], dtype=float32), array([0.25934333], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "3 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0344164],\n",
      "       [-3.3728302]], dtype=float32)>\n",
      "[array([[ 2.0344164],\n",
      "       [-3.3728302]], dtype=float32), array([0.26710516], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "4 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0361292],\n",
      "       [-3.3668377]], dtype=float32)>\n",
      "[array([[ 2.0361292],\n",
      "       [-3.3668377]], dtype=float32), array([0.2748011], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "5 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0191362],\n",
      "       [-3.3801608]], dtype=float32)>\n",
      "[array([[ 2.0191362],\n",
      "       [-3.3801608]], dtype=float32), array([0.28280982], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "6 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0148413],\n",
      "       [-3.3671145]], dtype=float32)>\n",
      "[array([[ 2.0148413],\n",
      "       [-3.3671145]], dtype=float32), array([0.29055855], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "7 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0011914],\n",
      "       [-3.3745432]], dtype=float32)>\n",
      "[array([[ 2.0011914],\n",
      "       [-3.3745432]], dtype=float32), array([0.29840845], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "8 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9960815],\n",
      "       [-3.346207 ]], dtype=float32)>\n",
      "[array([[ 1.9960815],\n",
      "       [-3.346207 ]], dtype=float32), array([0.30598342], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "9 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0026495],\n",
      "       [-3.3746135]], dtype=float32)>\n",
      "[array([[ 2.0026495],\n",
      "       [-3.3746135]], dtype=float32), array([0.31406182], dtype=float32)]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x.shape (1000, 2)\n",
      "10 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9969782],\n",
      "       [-3.3833625]], dtype=float32)>\n",
      "[array([[ 1.9969782],\n",
      "       [-3.3833625]], dtype=float32), array([0.32186714], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "11 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9873252],\n",
      "       [-3.3840227]], dtype=float32)>\n",
      "[array([[ 1.9873252],\n",
      "       [-3.3840227]], dtype=float32), array([0.32960436], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "12 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9867048],\n",
      "       [-3.3928826]], dtype=float32)>\n",
      "[array([[ 1.9867048],\n",
      "       [-3.3928826]], dtype=float32), array([0.33735818], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "13 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9895765],\n",
      "       [-3.3807497]], dtype=float32)>\n",
      "[array([[ 1.9895765],\n",
      "       [-3.3807497]], dtype=float32), array([0.3450627], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "14 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9841886],\n",
      "       [-3.3695788]], dtype=float32)>\n",
      "[array([[ 1.9841886],\n",
      "       [-3.3695788]], dtype=float32), array([0.3526792], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "15 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9640086],\n",
      "       [-3.3637774]], dtype=float32)>\n",
      "[array([[ 1.9640086],\n",
      "       [-3.3637774]], dtype=float32), array([0.36021766], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "16 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9697334],\n",
      "       [-3.3529198]], dtype=float32)>\n",
      "[array([[ 1.9697334],\n",
      "       [-3.3529198]], dtype=float32), array([0.3677394], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "17 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9719265],\n",
      "       [-3.3728733]], dtype=float32)>\n",
      "[array([[ 1.9719265],\n",
      "       [-3.3728733]], dtype=float32), array([0.37547627], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "18 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9503155],\n",
      "       [-3.3718572]], dtype=float32)>\n",
      "[array([[ 1.9503155],\n",
      "       [-3.3718572]], dtype=float32), array([0.38287485], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "19 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9725035],\n",
      "       [-3.3720603]], dtype=float32)>\n",
      "[array([[ 1.9725035],\n",
      "       [-3.3720603]], dtype=float32), array([0.39058843], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "20 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9778697],\n",
      "       [-3.3519335]], dtype=float32)>\n",
      "[array([[ 1.9778697],\n",
      "       [-3.3519335]], dtype=float32), array([0.39801848], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "21 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9988807],\n",
      "       [-3.3689775]], dtype=float32)>\n",
      "[array([[ 1.9988807],\n",
      "       [-3.3689775]], dtype=float32), array([0.40581548], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "22 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0140567],\n",
      "       [-3.379085 ]], dtype=float32)>\n",
      "[array([[ 2.0140567],\n",
      "       [-3.379085 ]], dtype=float32), array([0.41345537], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "23 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.017471 ],\n",
      "       [-3.3770876]], dtype=float32)>\n",
      "[array([[ 2.017471 ],\n",
      "       [-3.3770876]], dtype=float32), array([0.42097247], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "24 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9870392],\n",
      "       [-3.3784924]], dtype=float32)>\n",
      "[array([[ 1.9870392],\n",
      "       [-3.3784924]], dtype=float32), array([0.42864496], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "25 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9951557],\n",
      "       [-3.3921404]], dtype=float32)>\n",
      "[array([[ 1.9951557],\n",
      "       [-3.3921404]], dtype=float32), array([0.43625566], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "26 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9928461],\n",
      "       [-3.404561 ]], dtype=float32)>\n",
      "[array([[ 1.9928461],\n",
      "       [-3.404561 ]], dtype=float32), array([0.44380087], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "27 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9971559],\n",
      "       [-3.4065802]], dtype=float32)>\n",
      "[array([[ 1.9971559],\n",
      "       [-3.4065802]], dtype=float32), array([0.45131594], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "28 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0148141],\n",
      "       [-3.392317 ]], dtype=float32)>\n",
      "[array([[ 2.0148141],\n",
      "       [-3.392317 ]], dtype=float32), array([0.4588494], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "29 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0083961],\n",
      "       [-3.3650932]], dtype=float32)>\n",
      "[array([[ 2.0083961],\n",
      "       [-3.3650932]], dtype=float32), array([0.4662854], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "30 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0126014],\n",
      "       [-3.3601913]], dtype=float32)>\n",
      "[array([[ 2.0126014],\n",
      "       [-3.3601913]], dtype=float32), array([0.4736452], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "31 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.026435 ],\n",
      "       [-3.3765855]], dtype=float32)>\n",
      "[array([[ 2.026435 ],\n",
      "       [-3.3765855]], dtype=float32), array([0.48116234], dtype=float32)]\n",
      "\n",
      " epoch 1 , loss: 13.860939\n",
      "x.shape (1000, 2)\n",
      "0 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.043535],\n",
      "       [-3.368363]], dtype=float32)>\n",
      "[array([[ 2.043535],\n",
      "       [-3.368363]], dtype=float32), array([0.4883783], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "1 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.025613],\n",
      "       [-3.364383]], dtype=float32)>\n",
      "[array([[ 2.025613],\n",
      "       [-3.364383]], dtype=float32), array([0.49581614], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "2 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0096376],\n",
      "       [-3.3803499]], dtype=float32)>\n",
      "[array([[ 2.0096376],\n",
      "       [-3.3803499]], dtype=float32), array([0.5033958], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "3 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0164006],\n",
      "       [-3.3606305]], dtype=float32)>\n",
      "[array([[ 2.0164006],\n",
      "       [-3.3606305]], dtype=float32), array([0.5106252], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "4 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0239115],\n",
      "       [-3.382778 ]], dtype=float32)>\n",
      "[array([[ 2.0239115],\n",
      "       [-3.382778 ]], dtype=float32), array([0.51807517], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "5 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.020126 ],\n",
      "       [-3.3859847]], dtype=float32)>\n",
      "[array([[ 2.020126 ],\n",
      "       [-3.3859847]], dtype=float32), array([0.52542955], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "6 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0091074],\n",
      "       [-3.3673224]], dtype=float32)>\n",
      "[array([[ 2.0091074],\n",
      "       [-3.3673224]], dtype=float32), array([0.5327363], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "7 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.001144 ],\n",
      "       [-3.3758197]], dtype=float32)>\n",
      "[array([[ 2.001144 ],\n",
      "       [-3.3758197]], dtype=float32), array([0.5400893], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "8 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9956278],\n",
      "       [-3.3693807]], dtype=float32)>\n",
      "[array([[ 1.9956278],\n",
      "       [-3.3693807]], dtype=float32), array([0.5473323], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "9 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0036826],\n",
      "       [-3.3787193]], dtype=float32)>\n",
      "[array([[ 2.0036826],\n",
      "       [-3.3787193]], dtype=float32), array([0.5546878], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "10 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9814734],\n",
      "       [-3.3904514]], dtype=float32)>\n",
      "[array([[ 1.9814734],\n",
      "       [-3.3904514]], dtype=float32), array([0.5620458], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "11 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9898816],\n",
      "       [-3.389305 ]], dtype=float32)>\n",
      "[array([[ 1.9898816],\n",
      "       [-3.389305 ]], dtype=float32), array([0.56933886], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "12 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.982775 ],\n",
      "       [-3.4011054]], dtype=float32)>\n",
      "[array([[ 1.982775 ],\n",
      "       [-3.4011054]], dtype=float32), array([0.57660407], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "13 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.983949 ],\n",
      "       [-3.3735335]], dtype=float32)>\n",
      "[array([[ 1.983949 ],\n",
      "       [-3.3735335]], dtype=float32), array([0.583849], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "14 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9823607],\n",
      "       [-3.365079 ]], dtype=float32)>\n",
      "[array([[ 1.9823607],\n",
      "       [-3.365079 ]], dtype=float32), array([0.590977], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "15 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9577166],\n",
      "       [-3.3666086]], dtype=float32)>\n",
      "[array([[ 1.9577166],\n",
      "       [-3.3666086]], dtype=float32), array([0.59804654], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "16 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9826914],\n",
      "       [-3.3639278]], dtype=float32)>\n",
      "[array([[ 1.9826914],\n",
      "       [-3.3639278]], dtype=float32), array([0.6053826], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "17 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9725724],\n",
      "       [-3.3676808]], dtype=float32)>\n",
      "[array([[ 1.9725724],\n",
      "       [-3.3676808]], dtype=float32), array([0.6124517], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "18 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9490608],\n",
      "       [-3.3770952]], dtype=float32)>\n",
      "[array([[ 1.9490608],\n",
      "       [-3.3770952]], dtype=float32), array([0.6194446], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "19 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9704328],\n",
      "       [-3.3736188]], dtype=float32)>\n",
      "[array([[ 1.9704328],\n",
      "       [-3.3736188]], dtype=float32), array([0.62663853], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "20 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9675112],\n",
      "       [-3.3642652]], dtype=float32)>\n",
      "[array([[ 1.9675112],\n",
      "       [-3.3642652]], dtype=float32), array([0.6336268], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "21 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9932276],\n",
      "       [-3.3656957]], dtype=float32)>\n",
      "[array([[ 1.9932276],\n",
      "       [-3.3656957]], dtype=float32), array([0.6409045], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "22 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0181768],\n",
      "       [-3.3805523]], dtype=float32)>\n",
      "[array([[ 2.0181768],\n",
      "       [-3.3805523]], dtype=float32), array([0.64813745], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "23 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0181801],\n",
      "       [-3.3640313]], dtype=float32)>\n",
      "[array([[ 2.0181801],\n",
      "       [-3.3640313]], dtype=float32), array([0.6551236], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "24 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.009007 ],\n",
      "       [-3.3875136]], dtype=float32)>\n",
      "[array([[ 2.009007 ],\n",
      "       [-3.3875136]], dtype=float32), array([0.66242844], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "25 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.005389 ],\n",
      "       [-3.3928077]], dtype=float32)>\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[array([[ 2.005389 ],\n",
      "       [-3.3928077]], dtype=float32), array([0.66952103], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "26 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0093486],\n",
      "       [-3.407454 ]], dtype=float32)>\n",
      "[array([[ 2.0093486],\n",
      "       [-3.407454 ]], dtype=float32), array([0.6766004], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "27 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 1.9952488],\n",
      "       [-3.396721 ]], dtype=float32)>\n",
      "[array([[ 1.9952488],\n",
      "       [-3.396721 ]], dtype=float32), array([0.68369305], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "28 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0071065],\n",
      "       [-3.3840978]], dtype=float32)>\n",
      "[array([[ 2.0071065],\n",
      "       [-3.3840978]], dtype=float32), array([0.6907282], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "29 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0014942],\n",
      "       [-3.3695862]], dtype=float32)>\n",
      "[array([[ 2.0014942],\n",
      "       [-3.3695862]], dtype=float32), array([0.6976761], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "30 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.004658 ],\n",
      "       [-3.3679106]], dtype=float32)>\n",
      "[array([[ 2.004658 ],\n",
      "       [-3.3679106]], dtype=float32), array([0.70461994], dtype=float32)]\n",
      "x.shape (1000, 2)\n",
      "31 \n",
      " <tf.Variable 'sequential_8/dense_7/kernel:0' shape=(2, 1) dtype=float32, numpy=\n",
      "array([[ 2.0221405],\n",
      "       [-3.3519905]], dtype=float32)>\n",
      "[array([[ 2.0221405],\n",
      "       [-3.3519905]], dtype=float32), array([0.7113813], dtype=float32)]\n",
      "\n",
      " epoch 2 , loss: 12.250697\n"
     ]
    }
   ],
   "source": [
    "# 训练模型\n",
    "\n",
    "num_epoches = 3\n",
    "for epoch in range(num_epoches):\n",
    "    batch_data = db_all.shuffle(10)\n",
    "    batch_data = batch_data.batch(32)\n",
    "    for n, (train_x, train_y) in enumerate(batch_data):\n",
    "#         print(x, y)\n",
    "        print('x.shape', x.shape)\n",
    "        with tf.GradientTape() as tape:\n",
    "            l = loss(model(train_x), train_y)\n",
    "        # 自动记录可训练变量\n",
    "        grads = tape.gradient(l, model.trainable_variables)\n",
    "        # 完成梯度更新\n",
    "        trainer.apply_gradients(zip(grads, model.trainable_variables))\n",
    "        # model.trainable_variables查看模型参数\n",
    "        print(n , '\\n', model.trainable_variables[0])\n",
    "        # 通过 model.get_weights()[0]查看相关权重\n",
    "        print(model.get_weights())\n",
    "    # 每轮求一次全部的平均损失\n",
    "    l = loss(model(x), true_y)\n",
    "    print('\\n epoch %d , loss: %f' % (epoch, l.numpy()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 总结：\n",
    "１、创建模型：tf.kears.Sequential(), tf.keras.layer直接创建, 并且可以使用 tensorflow.initializers 进行参数初始化\n",
    "\n",
    "２、损失函数：tf.losses.MSE(x, y) 等各种，直接进行矩阵计算，得到平均损失，等价于：tf.reduce_mean(tf.square(predicted_y - tf.reshape(desired_y, predicted_y.shape)))\n",
    "\n",
    "３、优化算法（优化器）: tf.keras.optimizers.SGD(learning_rate=0.01)直接定义优化算法\n",
    "\n",
    "４、训练模型: tape.gradient(l, model.trainable_variables) 记录动态图梯度, trainer.apply_gradients(zip(grads, model.trainable_variables)) 进行梯度自动更新\n",
    "\n",
    "５、查看参数: model.trainable_variables 查看模型参数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (tf2-1)",
   "language": "python",
   "name": "tf2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}