CLEVRTask1SuperviseLearning.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reload_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "from tqdm import tqdm_notebook\n",
    "from source.datautil import *\n",
    "from source.model import *\n",
    "from source.trainutil import *\n",
    "DEVICE = \"cpu\" # or \"cuda:0\"\n",
    "# torch.cuda.set_device(DEVICE) # uncomment this for CUDA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CLEVR Dataset Prepare"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The first step is to download the CLEVR dataset from\n",
    "# https://cs.stanford.edu/people/jcjohns/clevr/\n",
    "# Download CLEVR v1.0 (18 GB) \n",
    "# Make sure the data utility can get image and mask without problem"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_path_json_train = \"dataset/json_clevr_train_with_mask.data\"\n",
    "PATH_TO_IMAGES_TRAIN = \"/storage/hezq17/CLEVR_v1.0/images/train\" # change this to where you download the images\n",
    "data_path_images_train = PATH_TO_IMAGES_TRAIN\n",
    "data_path_json_val = \"dataset/json_clevr_val_with_mask.data\"\n",
    "PATH_TO_IMAGES_VAL = \"/storage/hezq17/CLEVR_v1.0/images/val\" # change this to where you download the images\n",
    "data_path_images_val = PATH_TO_IMAGES_VAL"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data load from  dataset/json_clevr_train_with_mask.data\n",
      "Parsing dataset ...\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Test ClevrDataset utility\n",
    "clrdata = ClevrDataset(data_path_json_train, data_path_images_train)\n",
    "clrdata.get_mode=\"full\"\n",
    "rndpick = np.random.rand()*len(clrdata)\n",
    "img, mask_t, json_scene = clrdata[int(rndpick)]\n",
    "plt.imshow(img)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAV0AAADxCAYAAABoIWSWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAV6klEQVR4nO3df+xddX3H8eerLQgOXYUKK20znHZEXGJxTcE0WxDc5JdWEzE0myAh+7oEFoguAv6DLJKwZIKaOdwXYZQNwUYwYGU6hjTEhF8tVKTUhooMvrZr1wkFRvjR8t4f93Pp5dv7vd9z7z33nHPPeT2Sk9577uec8zlRXv30cz7n81FEYGZmxZhTdgXMzJrEoWtmViCHrplZgRy6ZmYFcuiamRXIoWtmViCHrpnZgCQdK2lTx/aCpIt7HuNxumZmw5M0F/gNcEJE/NdM5dzSNTPLxynAr3oFLsC8gipjZlYZ75Pi5Yxld8Bm4JWOXZMRMdml6NnALbOdz90LZtY4R0vx+YxlvwIbI2J5rzKSDga2Ax+IiJ29yrqla2aNI3LvWz0NeGS2wAWHrpk1kMg9/FaToWsB/CDNzBpqTsZtNpLeDvwZcHuW67qla2aNlFeLMyJeBo7IWt6ha2aNM4I+3cwcumbWSA5dM7OCuKVrZlawuSVd16FrZo0jHLpmZoVy94KZWUHcp2tmVjCHrplZQUbwGnBmDl0zayS3dM3MCqK0lWFkYS/pVElbJW2TdOmormNmNoi5Gbe8jSR001pB36I1x+RxwGpJx43iWmZm/WqPXshjlrF+jaqluwLYFhFPRcRrwK3AqhFdy8ysb2WF7qj6dBcBz3Z8nwJO6CwgaQKYaH2b+8dw2IiqYmb1smd3RLx7mDPUcfRCtz7qtyzGlhZ2mwSQ5gf8yYiqYmb1sq7nartZ1W30whSwpOP7YlqLtpmZla7MN9JGdd2HgaWS3pNWyTwbuHNE1zIz65sybnkbSUs3IvZKuhD4Ca1RFzdExOZRXMvMbBC1m2UsIu4C7hrV+c3MBlXHB2lmZpXlWcbMzApWtwdpZmaVltfLEZLmS/q+pF9K2iLpw73Ku6VrZo2Tc/fCN4AfR8Sn02itt/cq7NA1s0bKYziYpHcCfwp8DiBNe/Bar2PcvWBmjSPgoIwbsEDSho5touNUfwD8D/Avkh6V9B1Jv9Pr2g5dM2ukPvp0d0fE8o5tsuM084APAddGxPHA/wE9p7J16JpZ4+Q4teMUMBURD6bv36cVwjNy6JpZI+URuhHx38Czko5Nu04Bnuh1jB+kmVnj5Dx64W+Am9PIhaeA83oVduiaWSPlFboRsQlYnrW8Q9fMGsdzL5iZFcxzL5iZFcQT3piZFcyha2ZWIIeumVlB/CDNzKxgo1j/LAuHrpk1jhjTNdIkPQ28COwD9kbEckmHA98DjgGeBj4TEc8NV00zs3yN88oRH4mIZRHRfiPjUuCeiFgK3MMsM+6YmRUtxwlv+jaKc64C1qTPa4BPjuAaZmZDGdfQDeA/JG3smNj3qIjYAZD+PLLbgZIm2pMCzzLRuplZrtqjF7JseRv2nCsjYrukI4G7Jf0y64FpIuBJAGl+DFkPM7PMynwjbajrRsT29Ocu4AfACmCnpIUA6c9dw1bSzCxvZXUvDNzSTesAzYmIF9PnPwf+DrgTOBe4Kv15Rx4VNaui+1nX8/cPc2ZBNbF+jeMbaUcBP5DUPs93I+LHkh4G1ko6H3gGOGv4appVy2xhO72cw7daxnLCm4h4Cvhgl/3/S2vJCrNayRq0/R7rQC7H2IWuWRMME7TDXsNhPDqee8GsYooI237r4BDOV+oanV3kO7jKoWvWoQphO5P7WefgzYsE8zLG3+uv53ppL8FullQ5cNvGoY5jY968bFvel839jGZjyGHWMP20dHPm0LXGc+A20Jw5cMgh2cq++GLPn7vNttirvEPXzJon/5buRyJid5aCDl0zayZ3L5iZFaS/lu6C1myIb5pME3a1tWdbDOCfp/12AIeumTVPf6G7e5Z+2gNmW4yI+2Yq7NA1s+bJsU+3c7ZFSe3ZFh26ZnXglyNyImUfvdDzNDPOtjgjh66ZNU9+Ld2usy32OsCha2bNk1PozjTbYi8OXTNrHr+RZmZWIIeumVnBHLpmZgXpZ+6FnDl0zax5SuxemHU+XUk3SNol6fGOfYdLulvSk+nPd6X9kvRNSdskPSbpQ6OsvFmTeIxujtqhW9H5dG8E/hG4qWPfpcA9EXGVpEvT90uA04ClaTsBuDb9abXyVxnLXTfSWpgNrMoP0iLiPknHTNu9CjgpfV4DrKcVuquAmyIigAckzZe0MCJ25FVhK0vWoJ3pGAewVUxVQ3cGR7WDNCJ2pIkeABYBz3aUm0r7DghdSRPAROvboQNWw4oxSOCaVViJD9LyXiOt2/KaXZfSjIjJiFjemr3n4JyrYdXj4LYKqXifbjc7290GkhYCu9L+KWBJR7nFwPZhKmhmlrsqj16YwZ3AuenzucAdHfvPSaMYTgT2uD/XzCqpqi1dSbfQemi2QNIUcDlwFbBW0vnAM8BZqfhdwOnANuBl4Lzca2xmNqyKj15YPcNPp3QpG8AFw1bKzGykqhy6ZlY+vxiRM78GbGaDWrr0h7mc58knP57LecaGW7pmZgUZw9ELZlaQXl0LebVy8z5X5Y3hOF0zK4D7ckfED9LMbDoH7gjltBrwINy9YI33Yc6sXMCVVZ/GdDG4e8GsfNOD7n7WlV4HGxE/SDOrnqID0IFboJxbupLmSnpU0qx/U7ula9ZDUa1fB27B8m/pXgRsAd45W0GHrlkf8gxhB22JcgxdSYuBM4ArgS/MVt6haxlcx/Dz4dZz5YhBQ9iBW7J8Ry98HfgS8I4shd2naxkNE5r1DNxusoyEyCtwGzPSYBT669NdIGlDxzax/zQ6E9gVERuzXtotXetDOzyztHqbE7TduCVbcf11L+xurXDT1UrgE5JOBw4B3inp3yLiL2c6mUPXBtDsQLUayKlPNyIuAy5rnVInAX/bK3DBoWtjINb99Vu+68xvl1ST+mvUTGN+Ddhsv+lBO9NvTQ7gJ5/8uPt1BzWClyMiYj2wfrZyfpBmldMrcIcpa/am9iTmWba8L537Gc2GMEiIOnitbyXOvTBr6Eq6QdIuSY937PuKpN9I2pS20zt+u0zSNklbJX0s9xpbbQ0Tng5e61tVQxe4ETi1y/5rImJZ2u4CkHQccDbwgXTMP0mam1dlrb7yCE0Hr2VW5ZZuRNwH/Dbj+VYBt0bEqxHxa1pLsa8Yon7WAA5LK1yVQ7eHCyU9lrof3pX2LQKe7SgzlfYdQNJE+w0PeG2Iapjt17QAz3OIV6OGi43hg7RrgfcCy4AdwNfSfnUpG91OEBGTEbG89abHwQNWw8waFZZ5GqdJzCNiZ/uzpOvgzVk+poAlHUUXA9sHrp29RSzd2vcxevLYEdTEqsbB26dxWyNN0sKI2JG+fgpoj2y4E/iupKuBo4GlwEND17LBBgnabsc7fM06VDl0Jd0CnERrpp0p4HLgJEnLaHUdPA18HiAiNktaCzwB7AUuiIh9o6l6/Q0buNPP5eA1S6ocuhGxusvu63uUv5LWZL42hDwD18ymqXLomjXZ5RknJb/CUzmOlQh4bW85L+Q6dK10OvPbuQ31ymsCnKxh2628A7j6ImDv3nKu7dC1SsgzeAfRb8j2cy6HcPU4dM1yMGgrN8/AtfHg0DVjuNbuIIFbVNi2r+MWb7U4dG3kxmHYWL/BW+Wwnem6Dt/yuaVr1qEzSHsF8DgF7vQ6OHjL9cYb8Mor5VzboWuVludyPFUI3DYHb7nKbOl65QizklTpL4Em2rs325Y3t3St9qocbm7xlsMtXbMGq/JfCnXVDt0yWroO3YZp2pwO4xJo41LPumg/SMuy9SLpEEkPSfq5pM2Srpjt2u5eMLNGyqkV+ypwckS8JOkg4GeS/j0iHpjpALd0K6rq42nHwbi1HsetvuMsr+6FaHkpfT0obV1Xy2lzS9fMGifPB2lpxfONwPuAb0XEg73Ku6VrtTSurcZxrfe46bOlu6C9iG7aJt56rtgXEctoLU+2QtIf9bq2W7pm1jh9tnR3txbQne2c8byk9cCp7F/C7ABu6ZpVjFu7oxeR2+iFd0uanz4fCnwU+GWvY9zStdpxaNlscuzTXQisSf26c4C1EdHz/4BZFqZcAtwE/B7wBjAZEd+QdDjwPeAYWotTfiYinpMk4BvA6cDLwOci4pGBb8lyVfdREQ5cyyKv0I2Ix4Dj+zkmS/fCXuCLEfF+4ETgAknHAZcC90TEUuCe9B3gNFpLry8FJoBr+6mQ7Vf3gLSZ+S+P0SrzjbQsqwHvAHakzy9K2gIsAlbRWpodYA2wHrgk7b8pIgJ4QNJ8SQvTeawkDnCz/cZmPl1Jx9BqSj8IHNUO0ojYIenIVGwR8GzHYVNp31tCNw27SEMvDu274mZmw6h86Eo6DLgNuDgiXmh13XYv2mXfAW9oRMQkMNk69/yeb3A0WbuFOuicCW7hmh2o8pOYp3eKbwNujojb0+6d7W4DSQuBXWn/FLCk4/DFwPa8KtxUDk+z/FR6asc0GuF6YEtEXN3x053AuenzucAdHfvPUcuJwB7355pZlVT6QRqwEvgs8AtJm9K+LwNXAWslnQ88A5yVfruL1nCxbbSGjJ2Xa43NzHJQ2T7diPgZ3ftpAU7pUj6AC4asl5nZyIzN6AUzszqo/IM0s0H86Ec/zFTujDM+PuKamL2VW7pWeVkDdNBzO3itaJUdvWA2ysDNW11W1q3LfVSVF6a0yioqcPO8jgPLZlP1IWNmZrXiPl0ze5Nb6qPXnsS8DO5esFoa1+Aa13qPG3cvmJkVqNJzL5iNq3FrNY5bfceZW7pmZgVyS9dsRMal9Tgu9ayLvFYDHoRD12qv6oFW9frVkV+OMBsxB5t1cp+uWQGu4MxKrbLrvwjK4z5ds4I46Azya+lKWiLpXklbJG2WdNFs13ZL1xqn7Bavg78acmrp7gW+GBGPSHoHsFHS3RHxxEwHOHStkdrBV2T4OmyrI69JzNP6jzvS5xclbQEWATOGbpaFKbs2nyV9RdJvJG1K2+kdx1wmaZukrZI+NvSdmY1IUUHowK2WPrsXFkja0LFNdDunpGOA44EHe107S0u3a/M5/XZNRPzDtAsfB5wNfAA4GvhPSX8YEfsyXMuscKNu9Tpwq6fPB2m7I2J5rwKSDgNuAy6OiBd6lc2yMOVMzeeZrAJujYhXgV9L2gasAO6f7VpmZcozfB201ZfX6AVJB9EK3Jsj4vbZyvfVpzut+bwSuFDSOcAGWq3h52gF8gMdh03RO6TNKsWBWX95DRmTJOB6YEtEXJ3lmMxDxro0n68F3gsso9US/lq7aJfDo8v5Jtp9JPBa1mqYmQ0tx5cjVgKfBU7u9nyrm0wt3W7N54jY2fH7dfDmv8mmgCUdhy8Gtk8/Z0RMApOt4+cfEMpmZqOS4+iFn9G9oTmjLKMXujafJS3sKPYp4PH0+U7gbElvk/QeYCnwUD+VsurwKr1WV1V+DbjdfP6FpE1p35eB1ZKW0eo6eBr4PEBEbJa0ltY4tb3ABR65YGZVUuk10no0n+/qccyVwJVD1MvMbKQi3ijlun4jzcwaKIBy/gHu0DWzBgrKGjXl0DWzhnL3gplZQdy9YGZWIIeumVnBHLpmZgVxS9fMrEABvF7KlR26ZtZAbumamRXMoWsV5UlvrH7c0jUzK5hfjjAzK4hbumZmBfLcC2ZmBXJL18ysYO7TNTMriFu6ZmYFc+iamRWkvAdpWVYDPkTSQ5J+LmmzpCvS/vdIelDSk5K+J+ngtP9t6fu29Psxo70FM7N+Ba0+3Sxbb5JukLRL0uOzFiZD6AKvAidHxAeBZcCpkk4E/h64JiKWAs8B56fy5wPPRcT7gGtSOTOzitmXcZvVjcCpWa86a+hGy0vp60FpC+Bk4Ptp/xrgk+nzqvSd9PspkrqtJmxmVpL2g7ThQzci7gN+m/XKWVq6SJoraROwC7gb+BXwfES0V46fAhalz4uAZ1Nl9gJ7gCO6nHNC0gZJG8rqWzGzpuordBe0syptE8NcOdODtIjYByyTNB/4AfD+Ge4CoFurNg7YETEJTAJI8w/43cxstDKP090dEcvzumpfoxci4nlJ64ETgfmS5qXW7GJgeyo2BSwBpiTNA36XPpreZmajV+3RC+9OLVwkHQp8FNgC3At8OhU7F7gjfb4zfSf9/tOIcEvWzCokvz7dfmXp010I3CvpMeBh4O6IWAdcAnxB0jZafbbXp/LXA0ek/V8ALs291mZmQ8sndCXdAtwPHCtpStL5vcrP2r0QEY8Bx3fZ/xSwosv+V4CzZq2pmVlp2uN0czhTxOp+yvuNNDNrIM+9YGZWMIeumVlBPIm5mVmB8uvT7ZdD18waqtHdC3tegnVby67FCC0AdpddiRHy/Y23cbu/3x/+FH6QtjXP1+yqRtIG39/48v3VkUPXzKxAfpBmZlawZj9Imyy7AiPm+xtvvr/aKa97QZ6LxsyaRjo69i92M5uvbixtakczs3rwgzQzs4KV06ebabmeUZJ0qqStafXgsZwGsttqoJIOl3R3Wi35bknvSvsl6Zvpfh+T9KHyaj47SUsk3StpS1oN+qK0vy7314jVrtOSW49KWpe+1+r++vcGrdELWbZ8lRq6kuYC3wJOA44DVks6rsw6DehGDlwN9FLgnrRa8j3sn1f4NGBp2iaAawuq46D2Al+MiPfTWjHkgvS/UV3urymrXV9Ea/GBtrrd3wCqO4n5KK0AtkXEUxHxGnArrdWEx8oMq4F2roo8fbXkm9Iqyw/QWvZoYTE17V9E7IiIR9LnF2n9h7uI+txf7Ve7lrQYOAP4TvouanR/g6n2yhGj9ObKwUnnqsLj7qiI2AGt4AKOTPvH9p7TPzWPBx6kRvc3itWuK+brwJfY34l5BPW6vwG9kXHLV9mhm2nl4JoZy3uWdBhwG3BxRLzQq2iXfZW+v4jYFxHLaC2wuoIcVruuCklnArsiYmPn7i5Fx/L+Btfclm575eC2zlWFx93O9j+r05+70v6xu2dJB9EK3Jsj4va0uzb31xYRzwPr6VjtOv3UbbVrxmS165XAJyQ9Tav77mRaLd+63N8Qmhm6DwNL05PUg4Gzaa0mXAedqyJPXy35nPSU/0RgT/uf6VWU+vOuB7ZExNUdP9Xl/mq92nVEXBYRiyPiGFr/ff00Iv6Cmtzf4MobvVDqON2I2CvpQuAnwFzghojYXGadBpFWAz0JWCBpCrgcuApYm1YGfYb9i3XeBZwObANeBs4rvML9WQl8FvhF6vcE+DL1ub+FwJo0kmYOsDYi1kl6ArhV0leBR3nratf/qtZq17+lFWTj6BLqfX8Z+DVgM7NCSIcHfCxj6VtzfQ247O4FM7OS5NOn2+8LXg5dM2ugfEYvDPKCl+deMLMGCuD1PE705gteAJLaL3g9MdMBDl0za6A9P4EfLshY+BBJGzq+T0ZEew7ibi8DndDrZA5dM2uciJg+V8qg+n6ZxH26ZmaD6/tlIIeumdng+n7By90LZmYDGuQFL78cYWZWIHcvmJkVyKFrZlYgh66ZWYEcumZmBXLompkVyKFrZlYgh66ZWYH+H7fymb7JJMHpAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_mat(mask_t)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Task 1: Supervise Learning of Properties."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data load from  dataset/json_clevr_train_with_mask.data\n",
      "Parsing dataset ...\n",
      "Data load from  dataset/json_clevr_val_with_mask.data\n",
      "Parsing dataset ...\n"
     ]
    }
   ],
   "source": [
    "clrdata = ClevrDataset(data_path_json_train, data_path_images_train, get_mode=\"maskedimage_posicolormaterial\")\n",
    "clrdata_val = ClevrDataset(data_path_json_val, data_path_images_val, get_mode=\"maskedimage_posicolormaterial\")\n",
    "\n",
    "data_loader = torch.utils.data.DataLoader(\n",
    "    clrdata, batch_size=64, shuffle=True, num_workers=4, collate_fn=None)\n",
    "\n",
    "data_loader_val = torch.utils.data.DataLoader(\n",
    "    clrdata_val, batch_size=64, shuffle=True, num_workers=4, collate_fn=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Calculated Hsize: 6.0, Wsize: 11.0, ConvOut:2112.\n"
     ]
    }
   ],
   "source": [
    "model_para={\n",
    "    \"input_H\":320,\n",
    "    \"input_W\":480,\n",
    "    \"output_size\": 2+8+2, # posix, posiy, color, material\n",
    "    \"conv_para\":[[3,16,7,2],[16,32,5,1],[32,32,5,1],[32,32,5,1]],\n",
    "    \"maxpool_para\": 2,\n",
    "    \"mlp_para\": [128,64],\n",
    "    \"res_per_conv\":1\n",
    "}\n",
    "misc_para = {\n",
    "    \"postprocess_mode\":\"gaussian_noise\",\n",
    "    \"postprocess_partition\": [2,10,12],\n",
    "    \"noise_std\":0.2e-2\n",
    "}\n",
    "cnn = ConvNet(model_para, para=misc_para)\n",
    "\n",
    "misc_para = {\n",
    "    \"loss_flag\": \"posicolormaterial_loss\",\n",
    "}\n",
    "\n",
    "loss = CAL_LOSS(cnn,para=misc_para)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dict={\"train\":data_loader,\"val\":data_loader_val}\n",
    "run_para={\"loss_clip\":0.2}\n",
    "ptM = PyTrain_Main(loss, data_dict, device=DEVICE, para=run_para)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Adjust epoch number and learning rate until converge\n",
    "ptM.run_training(epoch=10, lr=1e-4, optimizer_label=\"adam\", print_step=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model load from  pretrained_models/CNN_posiColorMaterial_sqrt0p2%_gaussNoise.model\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "ConvNet(\n",
       "  (conv_layer_stack): ModuleList(\n",
       "    (0): Sequential(\n",
       "      (0): Conv2d(3, 16, kernel_size=(7, 7), stride=(2, 2))\n",
       "      (1): LayerNorm((157, 237), eps=1e-05, elementwise_affine=True)\n",
       "      (2): ReLU()\n",
       "      (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "    )\n",
       "    (1): Sequential(\n",
       "      (0): Conv2d(16, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "      (1): LayerNorm((74, 114), eps=1e-05, elementwise_affine=True)\n",
       "      (2): ReLU()\n",
       "      (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "    )\n",
       "    (2): Sequential(\n",
       "      (0): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "      (1): LayerNorm((33, 53), eps=1e-05, elementwise_affine=True)\n",
       "      (2): ReLU()\n",
       "      (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "    )\n",
       "    (3): Sequential(\n",
       "      (0): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "      (1): LayerNorm((12, 22), eps=1e-05, elementwise_affine=True)\n",
       "      (2): ReLU()\n",
       "      (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "    )\n",
       "  )\n",
       "  (res_conv_stack): ModuleList(\n",
       "    (0): Sequential(\n",
       "      (0): ResConvModule(\n",
       "        (conv): Conv2d(16, 16, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "        (lnorm): LayerNorm((78, 118), eps=1e-05, elementwise_affine=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "    (1): Sequential(\n",
       "      (0): ResConvModule(\n",
       "        (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "        (lnorm): LayerNorm((37, 57), eps=1e-05, elementwise_affine=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "    (2): Sequential(\n",
       "      (0): ResConvModule(\n",
       "        (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "        (lnorm): LayerNorm((16, 26), eps=1e-05, elementwise_affine=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "    (3): Sequential(\n",
       "      (0): ResConvModule(\n",
       "        (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "        (lnorm): LayerNorm((6, 11), eps=1e-05, elementwise_affine=True)\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "  )\n",
       "  (i2h): Linear(in_features=2112, out_features=128, bias=True)\n",
       "  (linear_layer_stack): ModuleList(\n",
       "    (0): Sequential(\n",
       "      (0): Linear(in_features=128, out_features=64, bias=True)\n",
       "      (1): LayerNorm((64,), eps=1e-05, elementwise_affine=True)\n",
       "      (2): ReLU()\n",
       "    )\n",
       "  )\n",
       "  (h2o): Linear(in_features=64, out_features=12, bias=True)\n",
       "  (gauss_noise): GaussNoise()\n",
       "  (softmax): LogSoftmax()\n",
       "  (gsoftmax): Gumbel_Softmax()\n",
       ")"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Or load a pretrained model\n",
    "load_model(cnn,\"pretrained_models/CNN_posiColorMaterial_sqrt0p2%_gaussNoise.model\",map_location=DEVICE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start evaluation ... 235\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "235it [05:37,  1.44s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluation Perplexity:  0.022089182181244203\n",
      "Time used in evaluation:  337.64008116722107\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "ptM.do_eval()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Auto-encoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Calculated Hsize: 6.0, Wsize: 11.0, ConvOut:2112.\n"
     ]
    }
   ],
   "source": [
    "model_para={\n",
    "    \"input_H\":320,\n",
    "    \"input_W\":480,\n",
    "    \"output_size\": 128, # posix, posiy, color, shape\n",
    "    \"conv_para\":[[3,16,7,2],[16,32,5,1],[32,32,5,1],[32,32,5,1]],\n",
    "    \"maxpool_para\": 2,\n",
    "    \"mlp_para\": [512,256],\n",
    "    \"res_per_conv\":1,\n",
    "    \"infobn_model\":{\"gs_head_dim\":2, \"gs_head_num\":64}\n",
    "}\n",
    "\n",
    "reg_lamda_scale = 64\n",
    "misc_para = {\n",
    "    \"postprocess_mode\": \"GSVIB\",\n",
    "    \"infobn_para\": {\"temp_scan_num\":4,\"reg_lamda\":0.008*reg_lamda_scale,\"scale_factor\":1.0}\n",
    "}\n",
    "\n",
    "cnn2 = ConvNet(model_para,para = misc_para)\n",
    "\n",
    "cnn.sfmsample_flag = True\n",
    "\n",
    "misc_para = {\n",
    "    \"cooprer_train_flag\": False,\n",
    "    \"coop_mode\": \"concatenate\"\n",
    "}\n",
    "encoder = ABS_CNN_COOP(cnn, cnn2, para=misc_para)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Calculated Hsize: 6.0, Wsize: 6.0, ConvOut:1152.\n",
      "Actual output shape (149.0,149.0).\n"
     ]
    }
   ],
   "source": [
    "model_para={\n",
    "    \"input_size\": 2+8+2+128, # posix, posiy, shape, material\n",
    "    \"output_H\":150,\n",
    "    \"output_W\":150,\n",
    "    \"deconv_para\":[[32,32,5,1], [32,16,5,1],[16,3,7,2]],\n",
    "    \"maxunpool_para\": 2,\n",
    "    \"mlp_para\": [256,512],\n",
    "    \"res_per_conv\": 1\n",
    "}\n",
    "misc_para = {}\n",
    "decnn = DeConvNet(model_para, para=misc_para)\n",
    "\n",
    "misc_para = {\n",
    "    \"seq1_coop_train_flag\": True\n",
    "}\n",
    "autoenc = ABS_CNN_SEQ(encoder, decnn, para=misc_para)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "misc_para={\"loss_flag\":\"mse_loss\"}\n",
    "loss = CAL_LOSS(autoenc,para=misc_para)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "clrdata.get_mode=\"auto_encode_focused\"\n",
    "clrdata_val.get_mode=\"auto_encode_focused\"\n",
    "\n",
    "data_loader = torch.utils.data.DataLoader(\n",
    "    clrdata, batch_size=64, shuffle=True, num_workers=4, collate_fn=None)\n",
    "\n",
    "data_loader_val = torch.utils.data.DataLoader(\n",
    "    clrdata_val, batch_size=64, shuffle=True, num_workers=4, collate_fn=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_dict={\"train\":data_loader,\"val\":data_loader_val}\n",
    "run_para={\"loss_clip\":20}\n",
    "ptM = PyTrain_Main(loss, data_dict, device=DEVICE, para=run_para)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Adjust epoch number and learning rate until converge\n",
    "# It takes pretty long time for this step\n",
    "ptM.run_training(epoch=100, lr=1e-4, optimizer_label=\"adam\", print_step=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model load from  pretrained_models/Autoenc_InfoB_PCM_10p53_h64reg0p5_pgnoise2e-3_focusMode_posiIncld.model\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "ABS_CNN_SEQ(\n",
       "  (seq1_coop): ABS_CNN_COOP(\n",
       "    (cooprer): ConvNet(\n",
       "      (conv_layer_stack): ModuleList(\n",
       "        (0): Sequential(\n",
       "          (0): Conv2d(3, 16, kernel_size=(7, 7), stride=(2, 2))\n",
       "          (1): LayerNorm((157, 237), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "          (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "        )\n",
       "        (1): Sequential(\n",
       "          (0): Conv2d(16, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "          (1): LayerNorm((74, 114), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "          (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "        )\n",
       "        (2): Sequential(\n",
       "          (0): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "          (1): LayerNorm((33, 53), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "          (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "        )\n",
       "        (3): Sequential(\n",
       "          (0): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "          (1): LayerNorm((12, 22), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "          (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "        )\n",
       "      )\n",
       "      (res_conv_stack): ModuleList(\n",
       "        (0): Sequential(\n",
       "          (0): ResConvModule(\n",
       "            (conv): Conv2d(16, 16, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "            (lnorm): LayerNorm((78, 118), eps=1e-05, elementwise_affine=True)\n",
       "            (relu): ReLU()\n",
       "          )\n",
       "        )\n",
       "        (1): Sequential(\n",
       "          (0): ResConvModule(\n",
       "            (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "            (lnorm): LayerNorm((37, 57), eps=1e-05, elementwise_affine=True)\n",
       "            (relu): ReLU()\n",
       "          )\n",
       "        )\n",
       "        (2): Sequential(\n",
       "          (0): ResConvModule(\n",
       "            (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "            (lnorm): LayerNorm((16, 26), eps=1e-05, elementwise_affine=True)\n",
       "            (relu): ReLU()\n",
       "          )\n",
       "        )\n",
       "        (3): Sequential(\n",
       "          (0): ResConvModule(\n",
       "            (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "            (lnorm): LayerNorm((6, 11), eps=1e-05, elementwise_affine=True)\n",
       "            (relu): ReLU()\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (i2h): Linear(in_features=2112, out_features=128, bias=True)\n",
       "      (linear_layer_stack): ModuleList(\n",
       "        (0): Sequential(\n",
       "          (0): Linear(in_features=128, out_features=64, bias=True)\n",
       "          (1): LayerNorm((64,), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (h2o): Linear(in_features=64, out_features=12, bias=True)\n",
       "      (gauss_noise): GaussNoise()\n",
       "      (softmax): LogSoftmax()\n",
       "      (gsoftmax): Gumbel_Softmax()\n",
       "    )\n",
       "    (trainer): ConvNet(\n",
       "      (conv_layer_stack): ModuleList(\n",
       "        (0): Sequential(\n",
       "          (0): Conv2d(3, 16, kernel_size=(7, 7), stride=(2, 2))\n",
       "          (1): LayerNorm((157, 237), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "          (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "        )\n",
       "        (1): Sequential(\n",
       "          (0): Conv2d(16, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "          (1): LayerNorm((74, 114), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "          (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "        )\n",
       "        (2): Sequential(\n",
       "          (0): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "          (1): LayerNorm((33, 53), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "          (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "        )\n",
       "        (3): Sequential(\n",
       "          (0): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "          (1): LayerNorm((12, 22), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "          (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
       "        )\n",
       "      )\n",
       "      (res_conv_stack): ModuleList(\n",
       "        (0): Sequential(\n",
       "          (0): ResConvModule(\n",
       "            (conv): Conv2d(16, 16, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "            (lnorm): LayerNorm((78, 118), eps=1e-05, elementwise_affine=True)\n",
       "            (relu): ReLU()\n",
       "          )\n",
       "        )\n",
       "        (1): Sequential(\n",
       "          (0): ResConvModule(\n",
       "            (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "            (lnorm): LayerNorm((37, 57), eps=1e-05, elementwise_affine=True)\n",
       "            (relu): ReLU()\n",
       "          )\n",
       "        )\n",
       "        (2): Sequential(\n",
       "          (0): ResConvModule(\n",
       "            (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "            (lnorm): LayerNorm((16, 26), eps=1e-05, elementwise_affine=True)\n",
       "            (relu): ReLU()\n",
       "          )\n",
       "        )\n",
       "        (3): Sequential(\n",
       "          (0): ResConvModule(\n",
       "            (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "            (lnorm): LayerNorm((6, 11), eps=1e-05, elementwise_affine=True)\n",
       "            (relu): ReLU()\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "      (i2h): Linear(in_features=2112, out_features=512, bias=True)\n",
       "      (linear_layer_stack): ModuleList(\n",
       "        (0): Sequential(\n",
       "          (0): Linear(in_features=512, out_features=256, bias=True)\n",
       "          (1): LayerNorm((256,), eps=1e-05, elementwise_affine=True)\n",
       "          (2): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (h2o): Linear(in_features=256, out_features=128, bias=True)\n",
       "      (infobn): GSVIB_InfoBottleNeck(\n",
       "        (softmax): LogSoftmax()\n",
       "        (gsoftmax): Gumbel_Softmax()\n",
       "      )\n",
       "      (softmax): LogSoftmax()\n",
       "      (gsoftmax): Gumbel_Softmax()\n",
       "    )\n",
       "  )\n",
       "  (seq2_train): DeConvNet(\n",
       "    (layer_norm0): LayerNorm((140,), eps=1e-05, elementwise_affine=True)\n",
       "    (i2h): Linear(in_features=140, out_features=256, bias=True)\n",
       "    (linear_layer_stack): ModuleList(\n",
       "      (0): Sequential(\n",
       "        (0): Linear(in_features=256, out_features=512, bias=True)\n",
       "        (1): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n",
       "        (2): ReLU()\n",
       "      )\n",
       "    )\n",
       "    (h2m): Linear(in_features=512, out_features=1152, bias=True)\n",
       "    (res_conv_stack): ModuleList(\n",
       "      (0): Sequential(\n",
       "        (0): ResConvModule(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "          (lnorm): LayerNorm((6, 6), eps=1e-05, elementwise_affine=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (1): Sequential(\n",
       "        (0): ResConvModule(\n",
       "          (conv): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "          (lnorm): LayerNorm((16, 16), eps=1e-05, elementwise_affine=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "      (2): Sequential(\n",
       "        (0): ResConvModule(\n",
       "          (conv): Conv2d(16, 16, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
       "          (lnorm): LayerNorm((36, 36), eps=1e-05, elementwise_affine=True)\n",
       "          (relu): ReLU()\n",
       "        )\n",
       "      )\n",
       "    )\n",
       "    (conv_layer_stack): ModuleList(\n",
       "      (0): Sequential(\n",
       "        (0): ConvTranspose2d(32, 32, kernel_size=(2, 2), stride=(2, 2))\n",
       "        (1): ConvTranspose2d(32, 32, kernel_size=(5, 5), stride=(1, 1))\n",
       "        (2): LayerNorm((16, 16), eps=1e-05, elementwise_affine=True)\n",
       "        (3): ReLU()\n",
       "      )\n",
       "      (1): Sequential(\n",
       "        (0): ConvTranspose2d(32, 32, kernel_size=(2, 2), stride=(2, 2))\n",
       "        (1): ConvTranspose2d(32, 16, kernel_size=(5, 5), stride=(1, 1))\n",
       "        (2): LayerNorm((36, 36), eps=1e-05, elementwise_affine=True)\n",
       "        (3): ReLU()\n",
       "      )\n",
       "      (2): Sequential(\n",
       "        (0): ConvTranspose2d(16, 16, kernel_size=(2, 2), stride=(2, 2))\n",
       "        (1): ConvTranspose2d(16, 3, kernel_size=(7, 7), stride=(2, 2))\n",
       "        (2): ReLU()\n",
       "      )\n",
       "    )\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "load_model(autoenc,\"pretrained_models/Autoenc_InfoB_PCM_10p53_h64reg0p5_pgnoise2e-3_focusMode_posiIncld.model\",map_location=DEVICE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Start evaluation ... 235\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "235it [12:32,  3.20s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Evaluation Perplexity:  10.472011829944368\n",
      "Time used in evaluation:  753.0462112426758\n"
     ]
    }
   ],
   "source": [
    "ptM.do_eval()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## AutoEncode Output Image Check"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "idp = int(np.random.rand()*15000)\n",
    "datax, labels = clrdata_val[idp]\n",
    "res = autoenc(datax.view(1,3,320,480).to(DEVICE), schedule=1.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "inimage = datax.permute(1,2,0)\n",
    "plt.imshow(inimage)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "outimage = autoenc.seq2_train.outimage.detach().cpu().squeeze().permute(1,2,0).numpy()\n",
    "plt.imshow(outimage)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Autoencoder pre-process to prepare for task2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "clrdata.get_mode=\"whole_pic_auto\"\n",
    "clrdata_val.get_mode=\"whole_pic_auto\"\n",
    "data_loader = torch.utils.data.DataLoader(\n",
    "    clrdata, batch_size=64, shuffle=True, num_workers=4, collate_fn=None)\n",
    "data_loader_val = torch.utils.data.DataLoader(\n",
    "    clrdata_val, batch_size=64, shuffle=True, num_workers=4, collate_fn=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "for ii in tqdm_notebook(range(len(clrdata_val))):\n",
    "    autoenc.auto_code_context_flag=True\n",
    "    loss = autoenc(clrdata_val[ii][0].to(DEVICE))\n",
    "    Nobj = len(clrdata_val.json_clevr[\"scenes\"][ii][\"objects\"])\n",
    "    for iio in range(Nobj):\n",
    "        contx = np.array(autoenc.auto_code_context[iio,:].detach().cpu())\n",
    "        clrdata_val.json_clevr[\"scenes\"][ii][\"objects\"][iio][\"auto_code\"]=contx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "save_data(clrdata_val.json_clevr,\"somedata_val.data\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "autoenc.to(DEVICE)\n",
    "for ii in tqdm_notebook(range(len(clrdata))):\n",
    "    autoenc.auto_code_context_flag=True\n",
    "    loss = autoenc(clrdata[ii][0].to(DEVICE))\n",
    "    Nobj = len(clrdata.json_clevr[\"scenes\"][ii][\"objects\"])\n",
    "    for iio in range(Nobj):\n",
    "        contx = np.array(autoenc.auto_code_context[iio,:].detach().cpu())\n",
    "        clrdata.json_clevr[\"scenes\"][ii][\"objects\"][iio][\"auto_code\"]=contx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "save_data(clrdata.json_clevr,\"somedata_train.data\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.5"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}