train_testing.ipynb

{
 "metadata": {
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  },
  "orig_nbformat": 2,
  "kernelspec": {
   "name": "python3",
   "display_name": "Python 3.8.5 64-bit ('TensorFlow-GPU': conda)"
  },
  "interpreter": {
   "hash": "7f846ec551663b66699b27a1e5b36a5620024784971716298b641ba6a3f6f1cb"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2,
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 13,
   "source": [
    "import config\r\n",
    "import game\r\n",
    "from visualise import vis\r\n",
    "import model\r\n",
    "\r\n",
    "import tensorflow.keras as keras\r\n",
    "\r\n",
    "import numpy as np\r\n",
    "from matplotlib import pyplot as plt"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "source": [
    "lines = open(\"training_data.log\").readlines()\r\n",
    "data = [[eval(x) for x in line.strip().split('\\t')]  for line in lines]\r\n",
    "\r\n",
    "xs = [l[0] for l in data]\r\n",
    "values = [float(l[1]) for l in data]\r\n",
    "probs = [l[2] for l in data]\r\n",
    "\r\n",
    "xs = np.array(xs)\r\n",
    "values = np.array(values)\r\n",
    "probs = np.array(probs)"
   ],
   "outputs": [],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "source": [
    "class MyModel():\r\n",
    "    def __init__(self):\r\n",
    "        inp_layer = keras.Input(shape=(9,))\r\n",
    "        l1 = keras.layers.Dense(20, activation=\"relu\")(inp_layer)\r\n",
    "        l2 = keras.layers.Dense(20, activation=\"relu\")(l1)\r\n",
    "        prob_logits = keras.layers.Dense(9, activation=\"relu\")(l2)\r\n",
    "        prob_head = keras.layers.Activation(\"softmax\")(prob_logits)\r\n",
    "        value_head = keras.layers.Dense(1, activation=\"tanh\")(l2)\r\n",
    "        self.model = keras.Model(inputs=inp_layer, outputs=[prob_head, value_head])\r\n",
    "        # self.pred_model = keras.Model(inputs=inp_layer, outputs=[prob_head, value_head])\r\n",
    "        self.model.compile(optimizer=keras.optimizers.SGD(learning_rate=0.1, decay=1e-4), loss=[\r\n",
    "            keras.losses.KLDivergence(),\r\n",
    "            keras.losses.MeanSquaredError()\r\n",
    "        ])\r\n",
    "\r\n",
    "    def predict(self, gamestate):\r\n",
    "        return self.model(gamestate.to_image())\r\n",
    "\r\n",
    "    def train(self, data, epochs, verbose = False):\r\n",
    "        xs, probs, values = data\r\n",
    "        history = self.model.fit(xs, [probs, values], batch_size=config.batch_size, epochs=epochs, verbose=verbose, shuffle=True)\r\n",
    "        return history\r\n",
    "\r\n",
    "net = MyModel()\r\n",
    "hist = net.train([xs, probs, values], 5000)\r\n",
    "plt.plot(hist.history['loss'])"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x266acf40b20>]"
      ]
     },
     "metadata": {},
     "execution_count": 34
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "source": [
    "hist = net.train([xs, probs, values], 5000)\r\n",
    "plt.plot(hist.history['loss'])"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x266acea2be0>]"
      ]
     },
     "metadata": {},
     "execution_count": 35
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ],
      "image/png": "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"
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "source": [
    "hist = net.train([xs, probs, values], 5000)\r\n",
    "plt.plot(hist.history['loss'])"
   ],
   "outputs": [
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x266af0c5610>]"
      ]
     },
     "metadata": {},
     "execution_count": 36
    },
    {
     "output_type": "display_data",
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ],
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAD4CAYAAADlwTGnAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/Il7ecAAAACXBIWXMAAAsTAAALEwEAmpwYAAAp2ElEQVR4nO3deXxV1bn/8c+TMAQQgkBAZTCgqICzEVDBgqBC8KodbivXodaBn7dqB7U1KliqoqhXq221XrQWxaq13qpcgyIiiAMIQVAZJUBkkCEgyiRDkuf3x9nhnqZATnL2yUnO+b5fr/PKOWvv9WQtXiFP9lp7r2XujoiIpJ+MZDdARESSQwlARCRNKQGIiKQpJQARkTSlBCAikqYaJbsBNdGuXTvPzc1NdjNERBqUuXPnbnL3nKrlDSoB5ObmUlRUlOxmiIg0KGb2xf7KNQQkIpKmlABERNKUEoCISJpSAhARSVNKACIiaUoJQEQkTSkBiIikqbRKAK/OW8uO3WXJboaISL2QNglg3qot/OJv8xn16oJkN0VEpF5IiwTQe8zbfPfxDwH4x7y1bNq+m5v+Np/yCm2GIyLpq0EtBVFbG7ft/qfPefe8DUDfo9ryw7zOyWiSiEjSpcUVwIFs3r4n2U0QEUmatE4A97+5JNlNEBFJmrROAAC5BYXc8/oi3v28NNlNERGpU2mfAACeen8lP356drKbISJSp5QAotzz+iIqdGeQiKQJJYAoT72/ku8/8SEL1n6T7KaIiCScEkAV81Z9zQV/eD/ZzRARSbi0SACn5x5a4zorN+1g1eadCWiNiEj9kBYJ4O/XnclZR7etUZ2B/zWdsx+cxsZtuxLUKhGR5IopAZjZEDNbambFZlawn+NdzGyamc0zs0/NLD/q2IlmNtPMFprZZ2aWFZRPD2LOD17tw+vWv/rrNX25ul/XGte77KmPqKhwtuzQQ2MiklqqTQBmlgk8BgwFegLDzaxnldNGAi+5+ynAJcDjQd1GwHPAde7eCxgA7I2qd6m7nxy8NsbbmeqMHNajxnU+37Cda54t4pS7p7D+G10NiEjqiOUKoDdQ7O4r3H0P8CJwUZVzHGgVvM8Gvgzenwd86u6fALj7Zncvj7/ZtWNm3DDwaG4dclyN6r2zJJKbNmxVAhCR1BFLAugIrI76vCYoizYauMzM1gCTgBuD8mMAN7PJZvaxmf26Sr2/BMM/o8zM9vfNzWyEmRWZWVFpafxP695y/rH854CjalX3osc+YM0WTQyLSGoIaxJ4ODDe3TsB+cAEM8sgstpoP+DS4Ot3zWxQUOdSdz8B6B+8Lt9fYHcf5+557p6Xk5MTUnPh8r5H1qpev/unkVtQSG5BIe8v2xRae0RE6losCWAtEL1mcqegLNrVwEsA7j4TyALaEblamOHum9x9J5Grg1OD89YGX7cBzxMZaqozd198PCVjh7FszFBevf4s7r6oV41jXPbnjxLQMhGRuhFLApgDdDezrmbWhMgk78Qq56wCBgGYWQ8iCaAUmAycYGbNgwnh7wCLzKyRmbULzm8MXAAkZauuxpkZnNy5NZefkcvHo85NRhNERJKi2gTg7mXADUR+mS8mcrfPQjO7y8wuDE67GbjWzD4BXgCu9IgtwMNEksh84GN3LwSaApPN7NOgfC3wZKg9q4U2LZrw3NV9alQnt6CQu19flKAWiYgkjrk3nMXP8vLyvKioKKHfY9uuvZww+q0a15t4w1kc2bYF2c0aJ6BVIiK1Z2Zz3T2vanlaPAlcEy2zGnN+rw50a9eiRvUu/OMHnPTbtyitsv2kiEh9pQSwH/99eR7v3DKgVnWHPjoj3MaIiCSIEkDINm3fQ25BIVMWbUh2U0REDkoJ4CBev7Ffrete+2xi5ypEROKlBHAQx3fM5oOCc5h/Z+1uD7130uKQWyQiEh4lgGp0bN2M1s2b1KruuBkr+HqnVhEVkfpJCaCGXr3+rBqdf/JdUxLUEhGR+CgBxOjze4bywrV9Oblz6xrX/clfZv9L2Serv9YtoyKSVEoAMWrSKIMzjorsKvbUFXkc37FVNTX+z7SlpeQWFPLUeyuofPDuosc+YNjv30tIW0VEYtEo2Q1oiAb37MCAY3OY+8UWlm3czshXY1vG6J7CxRye3YzTu0b2KN6oKwARSSJdAdRSo8wM+nRry2U1XFb6+uc/pveYqfs+P/NhCfNWbQm7eSIi1dJaQCHKLSisdd23b/oO7Q5pUus7jkREDkRrAdWBwT061L7uw+/qjiERqVNKACF64AcnAvDoJSfXOsY/Pl4TUmtERA5OQ0AJ8vaiDVwTx3IQV53VlTv/rWeILRKRdKUhoDo2uGcHTqrFMwOVnv5gZXiNERHZDyWABHr1p2fyyk/PrHX9peu3UbxxG2u27AyxVSIiETElADMbYmZLzazYzAr2c7yLmU0zs3lm9qmZ5UcdO9HMZprZQjP7zMyygvLTgs/FZvZ7M7PwulU/mBmndDmUz+8ZWqv65z8yg8EPz6Df/dP4dk95yK0TkXRXbQIws0zgMWAo0BMYbmZVB6dHEtkr+BQim8Y/HtRtBDwHXOfuvYABwN6gzp+Aa4HuwWtIvJ2pr5o0yuCDgnPiirGnrOKAx3bsLmPXXiUIEamZWK4AegPF7r7C3fcALwIXVTnHgcq1EbKBL4P35wGfuvsnAO6+2d3LzexwoJW7z/LILPSzwMXxdaV+69i6GSVjhzFv1LmMOLtbjev/cdoyAHaXlTNv1Ra27trLrr3llJVX0Os3kzlu1JthN1lEUlwsS0F0BFZHfV4D9KlyzmjgLTO7EWgBDA7KjwHczCYDOcCL7v5AEDP6fsc1Qdm/MLMRwAiALl26xNDc+u3QFk24Pb8Hkxeu54vNsY/tP/neSp58TxPDIhKesCaBhwPj3b0TkA9MMLMMIgmmH3Bp8PW7ZjaoJoHdfZy757l7Xk5OTkjNTb53fzWQXke0Ynjv8JJa8cbtvP7pl9WfKCJCbFcAa4HOUZ87BWXRriYYw3f3mcFEbzsif9nPcPdNAGY2CTiVyLxAp2piprzCn/UH4IXZq0KJN/jhdwG44MQjQoknIqktliuAOUB3M+tqZk2ITPJOrHLOKmAQgJn1ALKAUmAycIKZNQ8mhL8DLHL3dcBWM+sb3P1zBfBaKD1qgFbel0/xmKE8fumpocRbvG5rKHFEJLVVewXg7mVmdgORX+aZwNPuvtDM7gKK3H0icDPwpJn9ksiE8JXB5O4WM3uYSBJxYJK7V66Y9lNgPNAMeCN4pSUzo1GmMaTXYaHE2/rt3upPEpG0F9N+AO4+CZhUpezOqPeLgP3ulejuzxEZ8qlaXgQcX5PGprqMjHAehahoOKt7iEgSaUOYeqZk7DAAFn75DcN+/36tYvzj4zUMf3IWZ3Rrywsj+obZPBFJIVoKop7qdUQ2JWOHsfK+/OpPruLvcyN32M5csTnsZolIClECqOdScIUMEaknlAAagJX35bPk7tqtlDFh1hfs2F3Gw1M+Z2/5gZeTEJH0o/0AGpBVm3dy9oPT4ooxclgPrulf86UoRKThOtB+AEoADcyqzTt5r7iUO15ZEEq8Qce1589Xnh5KLBGpn7QhTIro0rY5PzitU/Unxmjqko2hxRKRhkUJoAFq2iiTZWOGUjxmKN8/tROX9olvPaHcgkJmfF4aUutEpKFQAmigGmdm0Cgzg4d+eBKXnB7/gnJXPD2biZ9oITmRdKIEkAJO6JQdSpyfvTCP3IJCHptWTLkeJxZJeUoAKeaULq3jjvHg5KVc8If3lQREUpwSQIp55EcnM7hHexbfFd8Om4vXbeWo2ydVf6KINFhaCyhF/PWaPmQ1zuDIti146seR2zqP7dCSpRu2xRX3qx17aNOiSRhNFJF6Rs8BpDB355kPSziq/SFc/ufZtY5TuUCdiDRMeg4gDZkZV57Vlf7dc+L6JZ5bUEhuQSHujrvz7MwSvt1THmJLRSQZNASURtq2aMLmHXtqXX/4k7OYteIrAJ6d+QVv3/Qd3J3ZK7+iT7e2YTVTROqIhoDSSHmFc9Ttk7imX1eeen9lKDFP7dKaj1d9zcBjc7h16HEcnt2M7GaNQ4ktIuGIawjIzIaY2VIzKzazgv0c72Jm08xsnpl9amb5QXmumX1rZvOD1xNRdaYHMSuPtY+ng1K9zAyjZOwwRl7Qc1/ZnDsG07dbm1rH/HjV1wBMW1rKkEfe45z/mh5nK0WkrlQ7BGRmmcBjwLnAGmCOmU0MtoGsNBJ4yd3/ZGY9iWwfmRscW+7uJx8g/KXB1pBSx47v2IqBx7Ynp2VT/vuyPIq++Iol67fx4OSlccXdvGMPuQWFLLl7CFmNM0NqrYgkQixzAL2BYndfAWBmLwIXAdEJwIFWwftsQGsK1HOv39h/3/vs5o0Z1KMDg3p04PqBR5NbUBh3/PvfXMJfPijh0OaNmXfneXHHE5HwxTIE1BFYHfV5TVAWbTRwmZmtIfLX/41Rx7oGQ0Pvmln/KvX+Egz/jLIDbH1lZiPMrMjMikpLtWBZQ/GXD0oA2LJzb3IbIiIHFNZtoMOB8e7eCcgHJphZBrAO6OLupwA3Ac+bWeWVwqXufgLQP3hdvr/A7j7O3fPcPS8nJyek5srBDO/dOdR4I54t4t5Ji3lnyQZKNu0INbaI1F4sCWAtEP0boVNQFu1q4CUAd58JZAHt3H23u28OyucCy4Fjgs9rg6/bgOeJDDVJPXDf904M9eGvtxZtYNyMFVw1vogBmiQWqTdiSQBzgO5m1tXMmgCXABOrnLMKGARgZj2IJIBSM8sJJpExs25Ad2CFmTUys3ZBeWPgAiCcLa4kNNNuGZDsJohIAlU7CezuZWZ2AzAZyASedveFZnYXUOTuE4GbgSfN7JdEJoSvdHc3s7OBu8xsL1ABXOfuX5lZC2By8Ms/E3gbeDIhPZRa69quBQCdDm3Gmi3fhhZ36uINlFU4g45rT1mF624hkSTRg2ASkz9MXcZDUz5nzHePD20/4krFY4bSKFOrkogkijaFl9Ds2F3Gl19/y7m/mxFazKvO6spN5x3DIU21OolI2LQYnISmRdNGdO/Qkjl3DObsY8K5M+vpD1Zy/G8mM3P5ZrbvLgslpogcnK4AJG4btu6iz71TQ425/N58MjP2+2iIiNSQrgAkYTq0yqJk7DCu6dc1tJhH3T6JjVt3hRZPRP6VBlwlNHcM60H7Vk1p06IpKzdtp0OrLO58bWGt4/W+dypTb/4OR+UcEmIrRaSSEoCExswYcfZR+z4vWPtN3DEHPfSudiQTSRANAUnCHN8xG4Dcts3jipNbUMiEWV+wa692IRMJkyaBJaFWf7WT7OaNWVm6gz3lFfz7EzPjiqdlpkVqTpPAkhSd2zSnVVZjTurcmtNz2zDh6viWfDpu1JtMXbwhpNaJpDddAUjSuDtdb5tUq7o/zOvEAz84KeQWiaQmXQFIvWNmvHzdGbWq+1LRGs64L9xnD0TSjRKAJFVebhsuPOmIWtVd980ucgsKyS0oZPG6rSG3TCT1aQhI6gV3Z9G6rXy04ivuen1R9RX2o0WTTE7v2obxP9HWEiLRNAQk9ZqZ0euIbK7q15XptwygW06LGsfYsaec6UtL2bhNTxCLxEIJQOqd3HYteOfmAXzym9ptJt97zFSmLdlIQ7q6FUkGJQCpt7KbNeaDgnNqVfcn4+dw+yufKQmIHIQSgNRrHVs3Y/otAzjzqLY1rvvC7NW1vs1UJB3ElADMbIiZLTWzYjMr2M/xLmY2zczmmdmnZpYflOea2bdmNj94PRFV5zQz+yyI+Xsz09q/sl+57Vrw/LV9a13/sWnFPPXeCl0NiFRRbQIINnV/DBgK9ASGm1nPKqeNBF5y91OIbBr/eNSx5e5+cvC6Lqr8T8C1RDaK7w4MqX03JB0svWcID3z/xBrXe3DyUu4pXMxHK79KQKtEGq5YrgB6A8XuvsLd9wAvAhdVOceBVsH7bODLgwU0s8OBVu4+yyN/lj0LXFyThkv6adookx+e3plZtw2qVf0n3l0ecotEGrZYEkBHYHXU5zVBWbTRwGVmtgaYBNwYdaxrMDT0rpn1j4q5ppqYAJjZCDMrMrOi0tLSGJorqe6w7CweveTkGtebvrSUO175jNyCQr7asSf8hok0MGHtBzAcGO/uD5nZGcAEMzseWAd0cffNZnYa8KqZ9apJYHcfB4yDyINgIbVXGrgLTzqCHoe3Ys2WnZgZP/nLnJjq/fWjVQCcevcUAG4+9xj+o08XWmY1xnGaNtJKo5I+YkkAa4HOUZ87BWXRriYYw3f3mWaWBbRz943A7qB8rpktB44J6neqJqbIAZkZx3RoyTEdWgLwxs/7s37rrpgTQaWHpnzOQ1M+3/dZm89IOollCGgO0N3MuppZEyKTvBOrnLMKGARgZj2ALKDUzHKCSWTMrBuRyd4V7r4O2GpmfYO7f64AXgulR5KWehzeioHHto87zpsL1vPl19+yddfeEFolUr9VewXg7mVmdgMwGcgEnnb3hWZ2F1Dk7hOBm4EnzeyXRCaEr3R3N7OzgbvMbC9QAVzn7pW3YvwUGA80A94IXiJxWTZmKGeOfYfSbbtrVf+65+buez9v1Lkc2qJJWE0TqXe0GJykpNyCwlDiDDg2h4f+/STaHtI0lHgiyaDF4CStlIwdxrxR58YdZ/rSUh6duown3l1ORUXD+WNJJBZKAJKyDm3RhMKf9Ys7zrMzv2DsG0v40bj49jMWqW+UACSl9ToimxvPOTqUWHNKttDrzjcBKN22W0tLSIOnBCAp7+bzjmXKL88OJdaOPeUMHzeL08e8zTMfloQSUyRZNAksaWPX3nKaNsrguY9WMerVBXHHa5XViK27yvjbiL706Vbz1UpF6oomgSXtZTXOxMy4vO+RzPjVwFotLBdt664yAH40bhYAL81Zzelj3qa8wtmxuyzu9ookmq4AJO2FccvolWfmMr7KkNCyMUNpnKm/sST5DnQFoAQgEsh/9D0WrdsaaszjDmvJm78IZ/5BpLY0BCRSjUk/78/3T40sUdWxdbNQYi5Zv43/mbuGcj1DIPWQEoBIlNEX9mTUBT15/9aBocW8+e+fcNTtka0pPyjexPLS7aHFFomHhoBEDqJ44zYGPzwjlFgXnHg4r3+6DtCqo1K3NAQkUgtHt28Z2i/ryl/+AFMXbwglpkg8lABEYtC/e7tQ4139TBG5BYX8Z9TqoyJ1LawdwURS2lM/zmPLjr0clp1F3j1T2LQ9nC0l31iwnimLNjBrxWau7d+Nw7KzQokrEgtdAYjEoGmjzH2/nJ+7pg+dDg3nLiGAa58t4s/vr6TvfVOZtWIzABUVzu6ycq03JAmlSWCRWnB3ut42iRsGHs33T+vEwP+aHmr8c45rzztLNvLdUzryux+dHGpsST96EEwkweav/po2zZuQkQH97p8WWlzdMSTxiusuIDMbYmZLzazYzAr2c7yLmU0zs3lm9qmZ5e/n+HYzuyWqrMTMPjOz+Wam3+rS4J3cuTVd2jan06HNQ42bW1DImwvW88CbS9hTVhFqbElv1SaAYFP3x4ChQE9guJn1rHLaSOAldz+FyKbxj1c5/jD73/N3oLufvL/MJNKQjf3eCTTOtNDiXffcXB6fvpwxhYtCiykSyxVAb6DY3Ve4+x7gReCiKuc40Cp4nw18WXnAzC4GVgIL426tSANxSe8uLBuTz4SrezPxhrNCi/vMzC/Ytbc8tHiS3mJJAB2B1VGf1wRl0UYDl5nZGmAScCOAmR0C3Ar8dj9xHXjLzOaa2YgDfXMzG2FmRWZWVFpaGkNzReqP/t1zOLFTa0rGDuPl687g3V8NiDvmcaPe5Ja/fxJ/4yTthXUb6HBgvLt3AvKBCWaWQSQx/M7d97f4ST93P5XI0NL1ZrbfJRPdfZy757l7Xk5OTkjNFal7ebltOLJti1BivTx3jfYckLjFkgDWAp2jPncKyqJdDbwE4O4zgSygHdAHeMDMSoBfALeb2Q3BeWuDrxuBV4gMNYmkjT8MPyWu+r1+M5kTfjOZpeu36XkBqZVYEsAcoLuZdTWzJkQmeSdWOWcVMAjAzHoQSQCl7t7f3XPdPRd4BLjX3f9oZi3MrGVwfgvgPCD+PfpEGoDCn/Xj4R+exL+ddATFY4Yy+45BtY61bXcZ5z8yg0feXqa5AamxahOAu5cBNwCTgcVE7vZZaGZ3mdmFwWk3A9ea2SfAC8CVfvA/SToA7wfnzwYK3f3NeDoi0lD0OiKb7wX7DjTKzKB9y6y4kgDAo1OXMfzJWWE0T9KIHgQTqWfi2aJy9L/1ZPT/LuKc49rz9JWnx1Tn8enFPPDmUt779UA6twn3GQapH/QksEgDsbusnF17Kzjpt2/FFeewVlncPqwHF550xEHPi044euo4NWk/AJEGommjTLKbNWbAsfHd9bZ+6y5+9sI8hjwyg9yCQn6lW0elCiUAkXrq8UtPDSXOkvXbAPj73DXc8vdPmLl8875jHxRvCuV7SMOk/QBE6qnmTRpRMnYY7s6qr3bynQenxx3z5blreHnuGlbcm48ZXPrUR/E3VBoszQGINBAlm3Yw+n8XMn1pYp+I79KmOTN+PTCh30PqluYARBq43HYtGP+T3gmfqF311c6Expf6QwlApAFafm8+K+/Lr/7EOOQWFHL/m0sS+j0kuZQARBqgzAzDzHj7prMZ3KND6PErbw390/TloceW+kMJQKQBO7p9S5684jTeS+CY/dkPhLe7mdQvugtIpIEzMzq3ac7ye/P5dm85TRtl0P2O/e2/VDurvtrJWwvXc16vw0KLKfWDrgBEUkRmhnFI00Y0zsxg3OWnhRp7xIS5nDh6cqgxJfmUAERS0Hm9DqNk7DDuuqgXAGcfE/9eGlt3af+BVKMhIJEUdnnfI7nopI5kN28c1yJzkpr0IJhImigrrwDg6DjnBzq2bsb7tw7ELLxN7yWxtBqoiABQXuHs2FNGq6zGFG/czuCH3+W2ocfRvEkmo15bGFOMdoc0oWjkuQluqYRFCUBEqlXTYaJuOS04tkNL/nRZuJPOEi4tBSEi1fqw4ByuPDM35vNXlO7gjQXrE9cgSaiYEoCZDTGzpWZWbGYF+znexcymmdk8M/vUzPL3c3y7md0Sa0wRqXtHtG7G6At7UTJ2GH/8j9g3rc8tKNw3xyANR7UJwMwygceAoUBPYLiZ9axy2kgiewWfQmTT+MerHH8Y2DfzFGNMEUmiC048okYLz8U7uSx1L5YrgN5AsbuvcPc9wIvARVXOcaBV8D4b+LLygJldDKwEomeXYokpIvXA8ntjX3TurYXr6XvvVBrS3GI6iyUBdARWR31eE5RFGw1cZmZrgEnAjQBmdghwK/DbWsQkiDHCzIrMrKi0NLHroIvIv8rMiP12zxET5rJ+6y4Wr9uWwBZJWMKaBB4OjHf3TkA+MMHMMogkht+5+/baBnb3ce6e5+55OTnxP80oIjU3/86a3fL57MySxDREQhVLAlgLdI763Ckoi3Y18BKAu88EsoB2QB/gATMrAX4B3G5mN8QYU0TqidbNm/BBwTkxn//inNXkFhTy2nz9t67PYkkAc4DuZtbVzJoQmeSdWOWcVcAgADPrQSQBlLp7f3fPdfdc4BHgXnf/Y4wxRaQe6di6GbNvH0TekYfGXOfnL87nyr/M5p0lGxLYMqmtahOAu5cBNwCTgcVE7vZZaGZ3mdmFwWk3A9ea2SfAC8CVfpBZoAPFjK8rIpJo7VtlMfKCmt2wN31pKVeNL2LnnjJ27S3nm517E9Q6qSk9CSwiNfbeslIyzJi2ZCNPvb+yxvWXjRlK40w9h1pX9CSwiISmf/cczjq6XY2vBip1v+MNNm7bFXKrpKaUAEQkLiVjh3FKl9Y1rtd7zFQqKhrOCEQqUgIQkbi9fN2ZtarX7fZJ5BYU4u58vGoL5UoIdUoJQETilplhcW1M/7c5q/ne4x/y/ybMDbFVUh0lABEJRec2zWu0dlC0gn98BsDbi3W7aF3SlpAiUq9U7klw/cCj+NX5xyW5NalNVwAiEqqHf3hSKHEem7acBWu/IbegkA1bdcdQIigBiEiovndqJ96/dSBnHtU27lgX/OF9AK748+y4Y8m/UgIQkdB1OrQ5T16RR/4Jh7Hgt+cz9ebvxBVv6QatLpoIehJYROrUB8WbGPXqAlZs2lGr+oe1ymLW7YNCblVq05PAIlIvnHV0O965ZUCt66/fuov8R9/jiqdn89GKzfvKv91Tzu6y8hBamD50F5CIJMWDPziRhV9uZfyHJTWuu2jdVlgHMz6PbBJ198XHM+rVBQB8UHAOGQaHZzcLs7kpSUNAIpJU67/ZxfOzV/H7qctCjVvbZxJSkYaARKReOiw7i5vOPYafDeoeatzcgsJ9zxQAVFQ4d762gOKNmlCupAQgIvXC8N6dObJtc96/tfZLSuzPlEUbuOf1RdxTuJhnZ37B4IdnhBq/IdMQkIjUO8UbtyX0F/X3TunI6It60SqrccK+R31yoCEgJQARqZfKK5znP/qCUa8lbrPAbu1aMKhHey7tcyS57Vok7PskW1wJwMyGAI8CmcBT7j62yvEuwDNA6+CcAnefZGa9gXGVpwGj3f2VoE4JsA0oB8r217iqlABE0s+HxZv4j6c+qrPvVzJ22L65g1TZuexACaDa20DNLBN4DDgXWAPMMbOJ7r4o6rSRRPb1/ZOZ9QQmAbnAAiDP3cvM7HDgEzP732BPYICB7r4prp6JSEo78+h2++7oWV66nUEPvZvQ7/fUeyv2vZ+yaAP5Jxye0O+XTLGktt5AsbuvcPc9wIvARVXOcaBV8D4b+BLA3XdG/bLPCs4TEamVo3IOoWTsMD4dfR6vXX9WQr7HPYWL971/5O3PE/I96otYEkBHYHXU5zVBWbTRwGVmtobIX/83Vh4wsz5mthD4DLguKiE48JaZzTWzEQf65mY2wsyKzKyotLQ0huaKSKprldWYkzq3ZuV9+cy6LXHLQny+YTtffv0tK0q3J+x7JFNYg1vDgfHu3gnIByaYWQaAu3/k7r2A04HbzCwrqNPP3U8FhgLXm9nZ+wvs7uPcPc/d83JyckJqroikAjPjsOwshiVwmObMse9wToKHnZIllgSwFugc9blTUBbtauAlAHefSWS4p130Ce6+GNgOHB98Xht83Qi8QmSoSUSkxh679FRKxg5j5LAejPnu8Qn7Pjv3lFV/UgMSSwKYA3Q3s65m1gS4BJhY5ZxVwCAAM+tBJAGUBnUaBeVHAscBJWbWwsxaBuUtgPOITBiLiNTaNf27cWmfIxMSO7egkJ53TuathetTJhFUexdQcAfPDcBkIrd4Pu3uC83sLqDI3ScCNwNPmtkviYztX+nubmb9gAIz2wtUAD91901m1g14xcwq2/C8u7+ZkB6KSNpZfm8+7s74D0voeUQr/uPJ8G4jHRFsXH9Cx2x6d23DqAt6hha7rulBMBFJeXvLK9hbXsHdry/ihdmr6dO1DXvLK/h41ddxxx7S6zCeuPw0Fq/bymGtsji0RZP4GxwyPQksIgKs2ryTLm2bA1D46ToAjm5/CM9/9AXPzPyiVjF/Mbg7j7wdWc307ouP54ITDq9XiUAJQESkBqJXEq2N+rQctZaDFhGpgTsb8Nh+rJQARET246p+XeOqn1tQyIAHp7G3vIIVpdvZtH13SC0Lj4aAREQO4Jtv9zLooels2r4nlHjJGhbSEJCISA1lN2vMy9edGVq89d/sYueeMhas/WbfBHQyaVN4EZGDODK4YwigaORg+t47lbKK2o2c9L1v6j99HnZicieKlQBERA7CzHj+mj60b5VFu0OaUnxvPk+9t4LpS0t5vzi+1ezdneCB2KRQAhARqcaZR//T0mZc078b1/Tvxqbtu1m75VtGTChiw9aaT/LOWvEVZxzVNqxm1pgmgUVEQrBh6y763Du1+hMPYNLP+lPhTtNGGXzz7V7yctuE1jY9CCYiUgc+XL4plLWHwrxjSHcBiYjUgdND+st93TffhhLnYJQARERC1Dgzg5KxwygZO4wJV9d+m5Orxhfxp+nLKSuvCLF1/0wJQEQkQfp3z2HC1b1pmVXz+20Wr9vK/W8u4eg73uDnL85LQOs0ByAiUmf2lFVwzMg3alU3njkBzQGIiCRZk0YZDDy2/uxtrucARETq0JNX5PHl17v4+ts9XPjHD5LalpiuAMxsiJktNbNiMyvYz/EuZjbNzOaZ2admlh+U9zaz+cHrEzP7bqwxRURSUaPMDLq0bc6JnVrXqF4ihuurTQBmlgk8BgwFegLDzazqQtkjgZfc/RQim8Y/HpQvAPLc/WRgCPDfZtYoxpgiIintvV8P5I78HvQ4vBWP/Ojkg55bXsv1hw4mliuA3kCxu69w9z3Ai8BFVc5xoFXwPhv4EsDdd7p7WVCeFZwXa0wRkZTWuU1zrj27G2/8vD8XnnTEQc8tT8AVQCxzAB2B1VGf1wB9qpwzGnjLzG4EWgCDKw+YWR/gaeBI4HJ3LzOzWGJW1h8BjADo0qVLDM0VEWl4MjKMOXcMZueeMrbs3MtNL81nRemOfceN8BeNC2sSeDgw3t0fMrMzgAlmdry7V7j7R0AvM+sBPGNmNboHyt3HAeMgchtoSO0VEal3clo2BZpyZFt45+YB7Cmr4KEpS7lh4NE0aRT+TZuxJIC1QOeoz52CsmhXExnjx91nmlkW0A7YWHmCuy82s+3A8THGFBFJa00aZXDb0B4Jix9LSpkDdDezrmbWhMgk78Qq56wCBgEEf+lnAaVBnUZB+ZHAcUBJjDFFRCSBqr0CCMbsbwAmA5nA0+6+0MzuAorcfSJwM/Ckmf2SyETvle7uZtYPKDCzvUAF8FN33wSwv5iJ6KCIiOyfloIQEUlxWgpCRET+iRKAiEiaUgIQEUlTSgAiImlKCUBEJE01qLuAzKwU+KKW1dsBm0JsTkOgPqeHdOtzuvUX4u/zke7+LxsRNKgEEA8zK9rfbVCpTH1OD+nW53TrLySuzxoCEhFJU0oAIiJpKp0SwLhkNyAJ1Of0kG59Trf+QoL6nDZzACIi8s/S6QpARESiKAGIiKSplE8AZjbEzJaaWbGZFSS7PfEws6fNbKOZLYgqa2NmU8xsWfD10KDczOz3Qb8/NbNTo+r8ODh/mZn9OBl9iZWZdTazaWa2yMwWmtnPg/KU7beZZZnZbDP7JOjzb4Pyrmb2UdC3vwV7aWBmTYPPxcHx3KhYtwXlS83s/CR1KSZmlmlm88zs9eBzSvcXwMxKzOwzM5tvZkVBWd39bLt7yr6I7DWwHOgGNAE+AXomu11x9Ods4FRgQVTZA0BB8L4AuD94nw+8ARjQF/goKG8DrAi+Hhq8PzTZfTtInw8HTg3etwQ+B3qmcr+Dth8SvG8MfBT05SXgkqD8CeA/g/c/BZ4I3l8C/C143zP4mW8KdA3+L2Qmu38H6fdNwPPA68HnlO5v0OYSoF2Vsjr72U71K4DeQLG7r3D3PcCLwEVJblOtufsM4KsqxRcBzwTvnwEujip/1iNmAa3N7HDgfGCKu3/l7luAKQTbedZH7r7O3T8O3m8DFgMdSeF+B23fHnxsHLwcOAd4OSiv2ufKf4uXgUFmZkH5i+6+291XAsVE/k/UO2bWCRgGPBV8NlK4v9Wos5/tVE8AHYHVUZ/XBGWppIO7rwverwc6BO8P1PcG+28SXOqfQuQv4pTudzAcMp/IvtpTiPw1+7W7lwWnRLd/X9+C498AbWlYfX4E+DWRnQMh0v5U7m8lB94ys7lmNiIoq7Of7Vg2hZcGwt3dzFLyvl4zOwT4H+AX7r418gdfRCr2293LgZPNrDXwCpH9tFOSmV0AbHT3uWY2IMnNqWv93H2tmbUHppjZkuiDif7ZTvUrgLVA56jPnYKyVLIhuAwk+LoxKD9Q3xvcv4mZNSbyy/+v7v6PoDjl+w3g7l8D04AziFzyV/7RFt3+fX0LjmcDm2k4fT4LuNDMSogM054DPErq9ncfd18bfN1IJNH3pg5/tlM9AcwBugd3EzQhMmE0McltCttEoHLW/8fAa1HlVwR3DvQFvgkuKycD55nZocHdBecFZfVSMLb7Z2Cxuz8cdShl+21mOcFf/phZM+BcInMf04AfBKdV7XPlv8UPgHc8Mjs4EbgkuGumK9AdmF0nnagBd7/N3Tu5ey6R/6PvuPulpGh/K5lZCzNrWfmeyM/kAuryZzvZs+CJfhGZOf+cyBjqHcluT5x9eQFYB+wlMs53NZGxz6nAMuBtoE1wrgGPBf3+DMiLinMVkQmyYuAnye5XNX3uR2Sc9FNgfvDKT+V+AycC84I+LwDuDMq7EfmFVgz8HWgalGcFn4uD492iYt0R/FssBYYmu28x9H0A/3cXUEr3N+jfJ8FrYeXvp7r82dZSECIiaSrVh4BEROQAlABERNKUEoCISJpSAhARSVNKACIiaUoJQEQkTSkBiIikqf8P+zEWYWtdjBgAAAAASUVORK5CYII="
     },
     "metadata": {
      "needs_background": "light"
     }
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "source": [
    "g = game.GameState()\r\n",
    "g = g.next_state(1)\r\n",
    "print(vis(str(g.to_image())))\r\n",
    "net.predict(g)"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      ". X .\n",
      ". . .\n",
      ". . .\n",
      "\n"
     ]
    },
    {
     "output_type": "execute_result",
     "data": {
      "text/plain": [
       "[<tf.Tensor: shape=(1, 9), dtype=float32, numpy=\n",
       " array([[0.17079718, 0.06823235, 0.1650302 , 0.1074262 , 0.10892744,\n",
       "         0.0891085 , 0.09356696, 0.05887019, 0.13804105]], dtype=float32)>,\n",
       " <tf.Tensor: shape=(1, 1), dtype=float32, numpy=array([[0.13537578]], dtype=float32)>]"
      ]
     },
     "metadata": {},
     "execution_count": 37
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "source": [
    "for x, v, p in zip(xs, values, probs):\r\n",
    "    if (x == g.to_image()).all():\r\n",
    "        print(x, v, p)"
   ],
   "outputs": [
    {
     "output_type": "stream",
     "name": "stdout",
     "text": [
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.10526316 0.         0.10526316 0.15789474 0.15789474 0.10526316\n",
      " 0.15789474 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] -1.0 [0.10526316 0.         0.10526316 0.15789474 0.15789474 0.10526316\n",
      " 0.15789474 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.10526316 0.         0.10526316 0.10526316 0.10526316 0.21052632\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.21052632 0.         0.10526316 0.10526316 0.10526316 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] -1.0 [0.10526316 0.         0.10526316 0.10526316 0.10526316 0.21052632\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.21052632 0.         0.10526316 0.10526316 0.10526316 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] -1.0 [0.10526316 0.         0.10526316 0.15789474 0.15789474 0.10526316\n",
      " 0.15789474 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.05263158 0.         0.05263158 0.63157895 0.05263158 0.05263158\n",
      " 0.05263158 0.05263158 0.05263158]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.21052632 0.         0.10526316 0.10526316 0.10526316 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.10526316 0.         0.10526316 0.10526316 0.10526316 0.21052632\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] -1.0 [0.10526316 0.         0.10526316 0.10526316 0.10526316 0.21052632\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.21052632 0.         0.10526316 0.10526316 0.10526316 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] -1.0 [0.10526316 0.         0.10526316 0.15789474 0.15789474 0.10526316\n",
      " 0.15789474 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] -1.0 [0.10526316 0.         0.10526316 0.10526316 0.10526316 0.21052632\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.21052632 0.         0.10526316 0.10526316 0.15789474 0.05263158\n",
      " 0.15789474 0.05263158 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.21052632 0.         0.10526316 0.10526316 0.15789474 0.05263158\n",
      " 0.15789474 0.05263158 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.21052632 0.         0.10526316 0.10526316 0.15789474 0.05263158\n",
      " 0.15789474 0.05263158 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.21052632 0.         0.10526316 0.10526316 0.15789474 0.05263158\n",
      " 0.15789474 0.05263158 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.21052632 0.         0.10526316 0.10526316 0.15789474 0.05263158\n",
      " 0.15789474 0.05263158 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] -1.0 [0.21052632 0.         0.10526316 0.10526316 0.15789474 0.05263158\n",
      " 0.15789474 0.05263158 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.21052632 0.         0.10526316 0.10526316 0.15789474 0.05263158\n",
      " 0.15789474 0.05263158 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] -1.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] -1.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.15789474 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.10526316]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.10526316 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.10526316 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.10526316 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.10526316 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.10526316 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 0.0 [0.10526316 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n",
      "[0 1 0 0 0 0 0 0 0] 1.0 [0.10526316 0.         0.15789474 0.10526316 0.15789474 0.10526316\n",
      " 0.10526316 0.10526316 0.15789474]\n"
     ]
    }
   ],
   "metadata": {}
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "source": [],
   "outputs": [],
   "metadata": {}
  }
 ]
}