5.3 多输入通道和多输出通道.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.0.0\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import os\n",
    "\n",
    "print(tf.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 一、多输入通道，单输出通道\n",
    "# 输入形状为 h * w * channel\n",
    "def corr2d(X, K):\n",
    "    c,h,w = K.shape\n",
    "    Y = tf.Variable(tf.zeros((X.shape[0] - h + 1, X.shape[1] - w +1, X.shape[2])))\n",
    "    Y_out = tf.Variable(tf.zeros((X.shape[0] - h + 1, X.shape[1] - w +1)))\n",
    "    for i in range(Y.shape[2]):\n",
    "        for j in range(Y.shape[1]):\n",
    "            for k in range(Y.shape[0]):\n",
    "                print(X[:,:,i])\n",
    "                # 注意这里切片方式，以 channel 通道方向切片,这里输出的 Y 也是 h * w * channel形式\n",
    "                Y[j, k,i].assign(tf.cast(tf.reduce_sum(X[j:j+h, k:k+w, i] * K[i,: ,:]), dtype=tf.float32))\n",
    "    # 得到的Y仍然是三个通道，然后求和合并成一个通道\n",
    "    Y_out = tf.reduce_sum(Y,axis=2)\n",
    "    return Y_out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X tf.Tensor(\n",
      "[[[2 2]\n",
      "  [2 8]\n",
      "  [4 1]\n",
      "  [8 1]]\n",
      "\n",
      " [[8 2]\n",
      "  [4 2]\n",
      "  [3 8]\n",
      "  [2 3]]\n",
      "\n",
      " [[2 9]\n",
      "  [5 5]\n",
      "  [1 4]\n",
      "  [8 5]]\n",
      "\n",
      " [[2 8]\n",
      "  [6 5]\n",
      "  [8 1]\n",
      "  [6 7]]], shape=(4, 4, 2), dtype=int32)\n",
      "K tf.Tensor(\n",
      "[[[1 0]\n",
      "  [0 1]]\n",
      "\n",
      " [[1 1]\n",
      "  [1 0]]], shape=(2, 2, 2), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 2 4 8]\n",
      " [8 4 3 2]\n",
      " [2 5 1 8]\n",
      " [2 6 8 6]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 2 4 8]\n",
      " [8 4 3 2]\n",
      " [2 5 1 8]\n",
      " [2 6 8 6]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 2 4 8]\n",
      " [8 4 3 2]\n",
      " [2 5 1 8]\n",
      " [2 6 8 6]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 2 4 8]\n",
      " [8 4 3 2]\n",
      " [2 5 1 8]\n",
      " [2 6 8 6]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 2 4 8]\n",
      " [8 4 3 2]\n",
      " [2 5 1 8]\n",
      " [2 6 8 6]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 2 4 8]\n",
      " [8 4 3 2]\n",
      " [2 5 1 8]\n",
      " [2 6 8 6]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 2 4 8]\n",
      " [8 4 3 2]\n",
      " [2 5 1 8]\n",
      " [2 6 8 6]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 2 4 8]\n",
      " [8 4 3 2]\n",
      " [2 5 1 8]\n",
      " [2 6 8 6]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 2 4 8]\n",
      " [8 4 3 2]\n",
      " [2 5 1 8]\n",
      " [2 6 8 6]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 8 1 1]\n",
      " [2 2 8 3]\n",
      " [9 5 4 5]\n",
      " [8 5 1 7]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 8 1 1]\n",
      " [2 2 8 3]\n",
      " [9 5 4 5]\n",
      " [8 5 1 7]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 8 1 1]\n",
      " [2 2 8 3]\n",
      " [9 5 4 5]\n",
      " [8 5 1 7]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 8 1 1]\n",
      " [2 2 8 3]\n",
      " [9 5 4 5]\n",
      " [8 5 1 7]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 8 1 1]\n",
      " [2 2 8 3]\n",
      " [9 5 4 5]\n",
      " [8 5 1 7]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 8 1 1]\n",
      " [2 2 8 3]\n",
      " [9 5 4 5]\n",
      " [8 5 1 7]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 8 1 1]\n",
      " [2 2 8 3]\n",
      " [9 5 4 5]\n",
      " [8 5 1 7]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 8 1 1]\n",
      " [2 2 8 3]\n",
      " [9 5 4 5]\n",
      " [8 5 1 7]], shape=(4, 4), dtype=int32)\n",
      "tf.Tensor(\n",
      "[[2 8 1 1]\n",
      " [2 2 8 3]\n",
      " [9 5 4 5]\n",
      " [8 5 1 7]], shape=(4, 4), dtype=int32)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tf.Tensor: id=5283, shape=(3, 3), dtype=float32, numpy=\n",
       "array([[18., 16., 16.],\n",
       "       [26., 20., 26.],\n",
       "       [30., 27., 17.]], dtype=float32)>"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = tf.random.uniform(shape=(4,4,2),maxval=10, minval=1, dtype='int32')\n",
    "print('X',X)\n",
    "K = tf.constant([[[1, 0],[0,1]],[[1, 1],[1,0]]])\n",
    "print('K',K)\n",
    "Y = corr2d(X, K)\n",
    "Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "def corr2d_multi_in_out(X, K):\n",
    "    return tf.stack([corr2d(X, k) for k in K],axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "def corr2d_multi_in_out_1x1(X, K):\n",
    "    h, w, c_i = X.shape\n",
    "    c_o = K.shape[0]\n",
    "    # X展开\n",
    "    X = tf.reshape(X,(h * w, c_i))\n",
    "    # K 展开\n",
    "    K = tf.reshape(K,(c_o, c_i))\n",
    "    Y = tf.matmul(K, tf.transpose(X))\n",
    "    return tf.reshape(Y, (c_o, h, w))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.65613544 0.9935862  0.41396034]\n",
      " [0.6595298  0.51581895 0.58577085]\n",
      " [0.66548884 0.4231608  0.758587  ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.83474994 0.81704736 0.9675939 ]\n",
      " [0.21148312 0.87913954 0.6991843 ]\n",
      " [0.6115159  0.54191315 0.1911031 ]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "tf.Tensor(\n",
      "[[0.6159954  0.57899964 0.66094935]\n",
      " [0.44979584 0.46862543 0.54280627]\n",
      " [0.9587865  0.5687027  0.87610555]], shape=(3, 3), dtype=float32)\n",
      "Y1 tf.Tensor(\n",
      "[[[1.0617476  1.2494551  0.98350066]\n",
      "  [0.7342243  0.9135824  0.9262045 ]\n",
      "  [1.1586474  0.7752186  1.0105582 ]]\n",
      "\n",
      " [[1.1089427  1.2760073  1.0640146 ]\n",
      "  [0.6911466  0.98919255 0.9612526 ]\n",
      "  [1.1392602  0.7918697  0.92221546]]], shape=(2, 3, 3), dtype=float32)\n",
      "Y2 tf.Tensor(\n",
      "[[[1.0617476  1.2494551  0.9835007 ]\n",
      "  [0.7342243  0.9135824  0.9262045 ]\n",
      "  [1.1586474  0.7752186  1.0105582 ]]\n",
      "\n",
      " [[1.1089427  1.2760074  1.0640146 ]\n",
      "  [0.6911466  0.98919255 0.9612526 ]\n",
      "  [1.1392602  0.79186976 0.92221546]]], shape=(2, 3, 3), dtype=float32)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tf.Tensor: id=7605, shape=(), dtype=float32, numpy=1.4600097e-07>"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X = tf.random.uniform((3,3,3))\n",
    "K = tf.random.uniform((2,3,1,1))\n",
    "\n",
    "Y1 = corr2d_multi_in_out_1x1(X, K)\n",
    "Y2 = corr2d_multi_in_out(X, K)\n",
    "print('Y1',Y1)\n",
    "print('Y2',Y2)\n",
    "tf.norm(Y1-Y2) < 1e-6\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (tf2)",
   "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": "3.6.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}